Fuel 223 (2018) 125–139

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Kinetic modeling and optimization of biodiesel production from white T

mustard (Sinapis alba L.) seed oil by quicklime-catalyzed transesterification
Milan D. Kostića, Ivica G. Djalovićb, Olivera S. Stamenkovića, Petar M. Mitrovićb,
⁎
Dušan S. Adamovićb, Mirko K. Kulinac, Vlada B. Veljkovića,
a
Faculty of Technology, University of Niš, 16000 Leskovac, Bulevar Oslobodjenja 124, Serbia
b
Institute of Field and Vegetable Crops, 21000 Novi Sad, Maksima Gorkog 30, Serbia
c
University of East Sarajevo, Faculty of Agriculture, Vuka Karadzića 30, Istočno Novo Sarajevo, Republic of Srpska, Bosnia and Herzegovina

A R T I C LE I N FO A B S T R A C T

Keywords: The biodiesel production from white mustard (Sinapis alba L.) seed oil (WMSO) by transesterification with
Biodiesel methanol over the quicklime powder was investigated in a batch stirred reactor. Two independent first-order
Kinetic modeling models with respect to triacylglycerols (TAGs) or a more complex model that combined the changing mechanism
Oil extraction and the first-order rate law with respect to TAGs and fatty acid methyl esters (FAMEs), respectively described
Optimization
successfully the kinetics of this transesterification reaction. Besides that, the response surface methodology
Sinapis alba
Transesterification
coupled with a full factorial design with replication was applied to model and optimize esters content with
methanol-to-WMSO molar ratio, catalyst amount and reaction time (X1, X2 and X3, respectively). The analysis of
variance indicated that all individual process factors, the interactions X1–X2 and X2–X3 and the quadratic term
X22 influenced significantly FAME content at the 95% confidence level. According to the reduced quadratic
model, complete conversion could be achieved with the catalyst loading of 9.8%–10% and the methanol-to-
WMSO molar ratio in the range between 6.1:1 and 11.6:1 in 50 min. WMSO was transesterified even faster than
sunflower oil in the presence of both quicklime and KOH, due to higher total content of unsaturated fatty acids.

1. Introduction
Tremendous attention has been paid to biodiesel by governments,
business sectors and scientific institutions all over the world in recent
years because of many positive technical characteristics and important
economic, environmental, social and political impacts [1]. Despite
these benefits, the main barrier to biodiesel commercialization is the
high price of its production, caused by the high cost of currently used
oily feedstocks (mainly edible vegetable oils), which makes 70–95% of
the total biodiesel cost [2]. Furthermore, even if the whole amount of
available edible oils is used for the biodiesel production, current diesel
requirements will not be satisfied [3]. Thus, other seed crops that could
grow on marginal lands and produce non-edible oils should be looked
for. Additional possibilities for the improvement of biodiesel produc-
tion are to use the heterogeneous catalysts, to optimize transester-
ification reactions, to use more effective reactors and to upgrade each
production stage.
White mustard (Sinapis alba L.), an annual plant of the family
Brassicaceae, is cultivated worldwide because of its numerous uses. The
aboveground parts are used in agriculture as a green manure and a
fodder crop as tasty young seedlings are edible [4]. White mustard seed
(WMS) has the largest agronomic value because of high oil and protein
contents and low starch content [5]. WMS oil (WMSO) contains mainly
oleic, linoleic, linolenic and erucic acid [6]. It is used in industry for
lightning and as lubricant [7] or diesel fuel additive [8], in traditional
medicine as anti-tumor, antiviral and analgesic agent [9], as well as in
food preparation as a condiment [10,11] and a preservative [12].
WMSO is usually extracted from ground seeds by the Soxhlet extraction
apparatus using n-hexane or petroleum ether, water or supercritical
CO2 extraction and cold pressing or expelling [13]. Press cake, a by-
product of oil recovery, can be used in poultry production [14].
WMSO is currently seen as a promising biodiesel resource [15,16].
Low quality WMSO has already been employed for biodiesel production
[17,18]. An overview of the previous studies of biodiesel production
from WMSO is given in Table 1. Mainly methanol and alkali hydroxides
were used in the biodiesel production from WMSO. NaOH was more
active than KOH as higher esters yield (92%) was achieved with the
former than with the latter (84%) [19]. The catalyst amount was in the
range between 0.3% and 1.8%, mostly about 1% of the oil weight while
the methanol-to-oil molar ratio was from 2:1 to 12:1, most frequently

⁎
Corresponding author.
E-mail address: veljkovicvb@yahoo.com (V.B. Veljković).

https://doi.org/10.1016/j.fuel.2018.03.023
Received 25 June 2017; Received in revised form 1 March 2018; Accepted 5 March 2018
Available online 20 March 2018
0016-2361/ © 2018 Elsevier Ltd. All rights reserved.
M.D. Kostić et al. Fuel 223 (2018) 125–139

6:1. The reaction temperature was usually close to the boiling point of
Reference methanol (60–65 °C). Methyl esters content lower than the prescribed

[27]

[22]
[15]

[28]
[19]

[21]

[25]
[24]

[26]

[23]
[8]
limit for biodiesel (96.5%) was most likely because of the use of crude
(unrefined) oil as feedstock. Exceptionally, Tabtabaei et al. [20,21]

Reaction optimization, esters

Properties as diesel additive

used the dewatered WMSO/water/tetrahydrofuran emulsion and me-

Esters yield and properties

Remark/objective of study

Emission characteristics of
thanol in the presence of NaOH to produce biodiesel while Issariyakul

Viscometric analysis of
Miscella and biodiesel
et al. [8] reported the production of biodiesel from WMSO by KOH-
Esters synthesis and

Fuel performances

Fuel performances
Reaction kinetics
catalyzed transesterification with methanol, ethanol, propanol and

Esters properties
butanol. A few studies are related to the process optimization using the

production
traditional “one-factor-at-a-time” method [19] and the kinetic mod-
properties

properties

biodiesel

blends
eling [22]. Table 1 indicates that no solid catalyst has been applied to
accelerate the WMSO transesterification. A number of recent studies are
related to fuel properties, performances and exhaust gas emission of
WMSO biodiesel and its blends with diesel fuel [11,16,23–26].
Yield (purity),

(99.3)/10 min
(> 98)/1.5 h;

64.75/1.5 h
CaO is frequently employed as a catalyst for transesterification of
92/75 min
84/75 min

96.56/2h
(85%)/–

(99)/4h

–/5 min
%/time

various feedstocks because of its high basicity, mild reaction condition,

(82)/2

–/2h
high esters yield, possible recycling, low cost and easy preparation from
–

natural or waste sources [29–31]. Since this reaction can be mass

Optimal conditions

transfer- or reaction rate-controlled, two independent first-order

6:1, 0.5%, 65 °C

models with respect to triacylglycerols (TAGs) [32] or a more complex

conditions

model that combines the changing mechanism and the first-order rate
Reaction

law with respect to TAGs and fatty acid methyl esters (FAMEs) [33],
8:1

60

respectively have been employed so far for describing the kinetics of

–

–
–

–
–

transesterification over CaO-based catalysts.

Temperature, °C

The biodiesel production from non-edible WMSO by transester-

ification with methanol over low-cost quicklime was investigated in a
40–60

50–75

Room

batch stirred reactor. The main goal was to select the better model
60

70

65

22
65

60

60

55

between the two above-mentioned models of Veljković et al. [32] and

–

Miladinović et al. [33] and to make it simpler and easier for applica-
CH3OK/0.2:1 catalyst-to-

NaOH/0.1–0.9% of oil

tion. In addition, the transesterification reaction was optimized using a

NaOH (150 mL, 1 M)
Catalyst/loading,%

full factorial design with replication in combination with the response

NaOH/0.8% of oil

NaOH/1.2% of oil
KOH/0.5% of oil

KOH/0.5% of oil
KOH/1.8% of oil

KOH/0.3% of oil
KOH/ 1% of oil

KOH/ 1% of oil

surface methodology in order to select the best reaction conditions

oil molar ratio

(methanol-to-WMSO molar ratio, catalyst amount and reaction time)

ensuring the maximum FAME content.
–

2. Theoretical background
12 wt% of KOH/ -
6:1, 7:1 and 8:1

Both kinetic modeling and statistical modeling and optimization of

Alcohol-to-oil

25:6 mL/mL
molar ratio

the transesterification of WMSO with methanol in the presence of

2:1–10:1

quicklime powder were applied in this study.

14:1
6:1
6:1

6:1

6:1

6:1

6:1

2.1. Kinetic modeling

Methanol + THF (1:1 mL/

For modeling the kinetics of the WMSO transesterification with

Ethanol, propanol, 1-

methanol over quicklime powder in a batch stirred reactor, two dif-

Type of alcohol

ferent reaction mechanisms were supposed: (1) the pseudo first-order

reaction in the heterogeneous and pseudo-homogeneous regimes and
Methanol

Methanol
Methanol

Methanol
Methanol

Methanol
Methanol

Methanol

Methanol

Methanol
butanol

(2) the changing mechanism combined with the TAG mass transfer
mL)

limitation. In both cases, the transesterification reaction presents the

stepwise conversion of TAGs via di- and monoacylglycerols (DAGs and
MAGs, respectively) to FAMEs and glycerol. Since the consumption of
Erlenmeyer flask/magnetic/
The review of the transesterification of Sinapis alba oil.

Stirred reactor, –/600 rpm

Stirred reactor, –/600 rpm

Stirred reactor, 500 mL/–
Type, volume of reactor/

DAGs and MAGs is faster than TAGs, the intermediates will not be
Stirred reactor, 500 mL/

Stirred reactor, 200 mL/

considered, so the WMSO transesterification is shown as the following

Stirred reactor, 2 L/–
agitation speed, rpm

overall reversible reaction:

Glass container/–

A + 3B ⇆ 3R + S (1)
magnetic
600 rpm

300 rpm

where A, B, R and S present TAGs, methanol, FAMEs and glycerol, re-

spectively.
–

Further assumptions were as follows:

Commercial mustard oil

Commercial mustard oil

Commercial mustard oil

Commercial mustard oil

Extracted mustard oil

Extracted mustard oil

Extracted mustard oil

Extracted mustard oil

Extracted mustard oil

Dehydrated oil/THF

a) The WMSO transesterification occurs via the initial heterogeneous

and later pseudo-homogeneous regimes where the mass transfer
miscella

resistance and chemical reaction, respectively control the overall

Mustard oil
Feed stock

reaction. As a consequence, the variation of FAME content with time

Table 1

is sigmoidal as it has been observed for the sunflower oil transes-

terification catalyzed by quicklime [33] or CaO [32].

126
M.D. Kostić et al. Fuel 223 (2018) 125–139

b) The reaction mixture is perfectly mixed, causing uniformity of the that includes the catalyst concentration. After integration, the following
composition and catalyst space distribution. final equation is obtained:
c) The transesterification occurs at the catalyst particles surface be-
−ln(1−xA) = kapp,1·t + C1 (5)
tween methoxide ions adsorbed on the active centers and TAG
molecules in the liquid phase close to the active centers. where C1 is the integration constant. Values of the two constants are
d) The mass transfer rate of methanol towards the catalytically active calculated from the slope and intercept of the linear dependence of
sites, the reverse reaction rate as well as the adsorption/desorption −ln(1−xA) on t , respectively.
rates of methanol, FAME and glycerol do not limit the overall pro-
cess rate. 2.1.2. Changing mechanism combined with TAG mass transfer limitation
e) The homogeneous catalysis is negligible since the catalyst amount is This model was originally developed by Miladinović et al. [33] for
higher than the critical value (≥1% of the oil weight) [34]. the transesterification of sunflower oil with methanol over CaO or
f) The intraparticle diffusion rate does not influence the transester- quicklime. The advantage of the proposed kinetic model, compared to
ification reaction rate since the catalyst is with small surface area the simpler irreversible first-order kinetic model, is its capability of
and low porosity [33]. describing the sigmoidal variation of FAME content with the progress of
g) The neutralization of free fatty acids is negligible because of their the reaction without complicated computations. This model involves
low content in the WMSO (< 1.0%). Also, the saponification reac- the changing mechanism and first-order reaction rate laws with respect
tion is also ignorable at the high employed methanol-to-WMSO to TAGs and FAMEs, respectively:
molar ratio, which drives the hydroxide-methoxide equilibrium to
methoxide formation, decreasing the mole fraction of hydroxide ⎛− dcA ⎞ = km· cA ·(cR0 + cR)
ions and hence, preventing the TAG saponification [35]. ⎝ dt ⎠ K + cA (6)
where km is the apparent reaction rate constant, cR0 is the hypothetic
According to several research groups, transesterification of various initial FAME concentration corresponding to the initial available active
feedstocks over different solid catalysts includes two pseudo-first order catalyst surface and K is the model parameter defining the TAG affinity
steps, i.e. the mass-transfer controlled regime, followed by the chemi- for the catalyst active sites. The apparent reaction rate constant is
cally controlled regime, called here heterogeneous and pseudo-homo- connected with the true reaction rate constant and the initial con-
geneous reaction regimes, respectively. This model has already been centrations of methanol and catalyst [33]:
verified for the transesterification of vegetable oils with methanol over
CaO [32], Ca(OH)2 [36], calcium methoxide [37] and CaO·ZnO [38]. km = k ·cB0·ccat (7)
Two independent pseudo-first order models, which express a shift in the where k is the true reaction rate constant while cB0 and ccat are the
rate-controlling step, are commonly used [32,36,37]. Other researchers initial concentrations of methanol and the catalyst. The model para-
use more complex models that include the mass transfer limitation meter K is also connected with the initial concentrations of methanol
describing the conversion during the whole reaction time only by a and catalyst [33]:
single kinetic expression [33,38–43].
K = K ′·cB0·ccat (8)
2.1.1. Pseudo first-order reaction in heterogeneous and pseudo- where K ′ is the “true” TAG affinity for the catalyst active sites.
homogeneous regimes If the concentrations of TAGs and FAMEs are expressed in term of
2.1.1.1. Heterogeneous reaction regime. According to the assumption (a), the TAG conversion degree, Eq. (6) can be transformed as follows:
the TAG mass transfer resistance controls the overall reaction rate in the
dxA (1−xA)·(cR0 + 3·cA0·xA )
heterogeneous reaction regime occurring in the initial period of the = km
dt K + cA0 (1−xA) (9)
reaction. Therefore, the TAG disappearance rate must be equal to the
TAG mass transfer rate from the oil phase into methanol drops through This model has been proved for the transesterification of sunflower
the interface. Considering the assumption (b) of perfect mixing, the oil with methanol or ethanol over several calcium-based catalysts
kinetic equation can be written as [32]: [33,40,42,43], as well as for the transesterification of waste lard with
dcA methanol over CaO and quicklime [41]. In these studies, the kinetic
(−rA ) = − = kc a (cA−cA,s ) parameters were calculated from the above differential equation by the
dt (2)
nonlinear regression method.
where kc is the TAG mass transfer coefficient, a is the average specific Eq. (9) can further be transformed as follows:
interfacial area, cA and cA,s are the TAG concentrations in the oil phase
and on the interfacial area, respectively and t is the reaction time. As
cA,s = 0 and cA = cA0 (1−xA) , where cA0 is the initial TAG concentration,
dxA
= 3km
(1−xA)· 3cR0 + xA
A0 ( c
)
K
dt + 1−xA
then after integration, the following equation is obtained [32]: cA0 (10)

−ln(1−xA) = kc a·t (3) If 3cA0 ≫ cR0 , Eq. (10) is simplified into the following differential
equation:
where xA is the TAG conversion degree and kc a is the volumetric TAG
mass transfer coefficient, which can be calculated from the slope of the dxA (1−x )·x
= 3km K A A
linear dependence of −ln(1−xA) on t . dt + 1−xA
c A0 (11)

2.1.1.2. Pseudo-homogeneous reaction regime. In the pseudo- which can be solved analytically:
homogeneous regime, the chemical reaction is slower than the TAG xA(1 + cA0/ K )
mass transfer and limits the overall reaction rate. Assuming the ln = 3kapp,2 cA0 t + C1
(1−xA) (12)
irreversible pseudo-first order rate law, the kinetic equation is as
follows [32]: where kapp,2 = km/ K is the apparent reaction rate constant and C1 is the
integration constant. Small values of cR0 has already been reported for
dcA
(−rA) = − = kapp,1·cA the transesterification of sunflower oil over quicklime conducted under
dt (4)
the similar reaction conditions (< 0.002 mol/L) [43]. Under the ap-
where kapp,1 is the apparent pseudo-first order reaction rate constant plied reaction conditions, 3 cA0 is 2.0–2.4 mol/L, thus proving the

127
M.D. Kostić et al.
validity of Eqs. (11) and (12). Eq. (12) is valid for 0 < xA < 1.
If K ≫ cA , which is valid only for the later phase of the transester-
ification over quicklime, Eq. (11) is transformed as follows:
dxA 3k c
= m A0 (1−xA) xA
dt K (13)
For the transesterification of sunflower oil over quicklime con-
ducted under the similar reaction conditions, the K-values of
1.1–6.0 mol/L were reported [43] while cA0 is 0.67–0.81 mol/L for the
initial methanol-to-oil molar ratio in the range between 6:1 and 12:1.
Hence, the assumption of K ≫ cA might be questionable in the begin-
ning of the reaction, especially for the lowest catalyst concentration
(2% of oil), where the parameter K had the lowest value (close to 1.0)
and the initial TAG concentration was the largest (0.81 mol/L). How-
ever, the TAG concentration decreased rapidly in the medium and final
phases of the reaction, where the assumption of K ≫ cA was fulfilled, as
it will be shown.
After integration of Eq. (13), the following equation is derived:
xA
ln = 3kapp,2 cA0 t + C2
1−xA
where kapp,2 = km/ K is the apparent reaction rate constant and C2 is the
integration constant. Hence, the apparent reaction rate constant, kapp,2 ,
can be calculated as the slope of the dependence of ln[xA /(1−xA)] versus
time by the linear regression method. On the basis of Eqs. (7) and (8),
the apparent reaction rate constant is kapp,2 = k / K ′, which means that it
does not depend on the catalyst and methanol concentrations.
2.2. Statistical modeling and optimization of WMSO transesterification
The quadratic equation was used to connect FAME content with
methanol-to-WMSO molar ratio, catalyst loading and reaction time:
Y = b0 + b1 X1 + b2 X2 + b3 X3 + b12 X1 X2 + b13 X1 X3 + b23 X2 X3 + b11 X12
+ b22 X22 + b33 X32
where Y is the FAME content (dependent variable or response), b0 is the
constant regression coefficient, bi and bii are the linear and quadratic
regression coefficients, respectively and bij are the regression coeffi-
cients of two-factor interactions (i, j = 1, 2, 3) while X1, X2 and X3 are
the process factors (independent variables, i.e. methanol-to-WMSO
molar ratio, catalyst loading and reaction time, respectively). Data
processing and evaluating were performed using the Design Expert
software (Stat-Ease Inc., Minneapolis, USA). This software uses multiple
nonlinear regression to determine the parameters of Eq. (15) while its
fit was evaluated by the analysis of variance (ANOVA) revealing the
statistically significant process factors and their interactions with a
confidence level of 95% (p-value < 0.05). The statistical significance of
the developed model and the lack of fit were also evaluated based on
their F- and p-values. The developed quadratic equation was then
modified by eliminating the statistically insignificant terms. A number
of optimization points, where the maximum FAME content was
achieved for a set of the reaction conditions, were also found by the
same computer program.
3. Experimental
3.1. Materials
3.1.1. S. alba seeds
The seeds of the “NS Bela” variety of S. alba L. (WMS), created at the
Institute of Field and Vegetable Crops (Novi Sad, Serbia), were used. Oil
and moisture contents of the seeds were 20.64 ± 0.18 g/100 g and
3.78 ± 0.16 g/100 g and, respectively [13]. The WMSO was obtained
by cold pressing through an oil press (Komet, Germany) using 8 mm
nozzles. After pressing, WMSO was filtered under vacuum to remove
solid residues. The low content of free fatty acids (1.95 ± 0.03 mg
Fuel 223 (2018) 125–139
KOH/g) allowed the application of direct base-catalyzed transester-
ification for biodiesel synthesis. The WMSO consisted of palmitic
(0.73%), stearic (0.30), oleic (13.95%), arachidic (0.28%), eicosenoic
(7.41%), behenic (0.39%), erucic (59.98%), lignoceric (0.30%) and
nervonic (2.95%) acid [13].
3.1.2. Catalyst preparation
The raw quicklime was purchased at a local market. It was crushed,
milled and sieved through a set of standard sieves. The fraction passed
through the 0.5 mm sieve was calcined at 550 °C within 4 h to obtain
the powdered catalyst [33]. The important properties of the powdered
quicklime can be found elsewhere [33]. After cooling in a desiccator
containing CaCl2 and KOH, the calcined quicklime powder was stored
in a dark, well-closed glass bottle, which was kept in the desiccator.
3.1.3. Other chemicals
Methanol (99.5%) and KOH (pellets; 85%) were from Zorka-Pharma
(Serbia) and Moslab (Serbia), respectively. HCl (36%) was purchased
from Centrohem (Serbia), while HPLC grade n-hexane, 2-propanol and
(14)
methanol were purchased from Lab-Scan (Ireland). The HPLC standards
for methyl esters and triacylglycerols were obtained from Sigma Aldrich
(USA).
3.2. Transesterification of WMSO
3.2.1. Equipment and experimental procedure
A 250 mL glass three-neck round-bottom flask, connected to a
condenser, was used as a reactor for both homogeneous and hetero-
geneous transesterification of WMSO, which were performed using
dissolved KOH and powdered quicklime as catalysts, respectively. The
reactor was placed in a bath in order to maintain the reaction tem-
perature at 60.0 ± 0.1 °C. The reaction mixture was vigorously agi-
tated by a magnetic stirrer at the speed of 400 or 900 rpm for homo-
(15) geneous and heterogeneous catalysis, respectively that ensured the
uniform reaction mixture. The reactor was charged with the desired
amounts of methanol and the catalyst and heated to the required
temperature under stirring. The WMSO was preheated separately to the
same temperature and then added to the reactor when the reaction was
timed on. Samples of the reaction mixture were taken with the progress
of the reaction, neutralized instantly by the required amount of a HCl
solution and centrifuged at 3500 rpm (average 700 × g) for 10 min. The
ester layer was analyzed by the HPLC method described elsewhere [42].
3.2.2. Full factorial design for transesterification of WMSO over quicklime
A 33 full factorial design of experiments with repetition was em-
ployed for statistical modeling and optimization of FAME content
achieved in the transesterification of WMSO with methanol over
quicklime. The transesterification reaction was carried out in duplicate
for each set of the process factors’ levels in order to estimate the pure
error associated with repetitions (54 runs in total). For minimizing er-
rors, the experiments were conducted randomly. The experimental
design included three factors, namely methanol-to-WMSO molar ratio
(X1), catalyst loading (X2) and reaction time (X3) each at three levels
(Table 2). The ranges of the process factors were selected on the basis of
the previous work in the field of the transesterification of vegetable oils
Table 2
Experimental range and levels of each process factor.
Factor Unit Low level Middle level High level
(−1) (0) (+1)
Methanol-to-WMSO molar mol/mol 6:1 9:1 12:1
ratio (X1)
Catalyst loading (X2) % of oil 2 6 10
Reaction time (X3) min 30 40 50
128
M.D. Kostić et al. Fuel 223 (2018) 125–139

with methanol over CaO [30,31]. So far, these reactions were con- 3.6. Statistical analysis
ducted at the methanol-to-oil molar ratio of 3:1–21:1 (most frequently
6:1–12:1) and the catalyst loading of 1–10% (of oil). When the reaction The performances of the models developed were statistically esti-
is conducted under atmospheric pressure, the reaction temperature mated by the mean relative percent deviation (MRPD) and the coeffi-
close to the boiling point of methanol is commonly applied. cient of determination (R2), which were computed using the following
equations, respectively:
n
3.3. Crude biodiesel purification 100 xAp,i−xAa,i
MRPD =
n
∑ xAa,i
i=1 (16)
At the end of the homogeneous WMSO transesterification, the FAME
fraction (crude biodiesel) was gravitationally separated from the me- and
thanol-glycerol fraction in a separation funnel. Crude biodiesel was n
washed with distilled water (10% of the biodiesel weight) while stirred ∑ (xAp,i−xAa,i )2
with a magnetic stirrer (500 rpm) for 1 h. After separating water by R2 = i=1
n
centrifugation at 3500 rpm (average 700 × g) for 10 min, the washed
biodiesel was dried with anhydrous sodium sulfate, which was sepa-
∑ (xAp,i−xAm)2
i=1 (17)
rated by filtration.
The final reaction mixture of the heterogeneous transesterification where xA,p,i and xA,a,i are the predicted and actual values of TAG con-
was allowed to separate the solid and liquid phase. The liquid phase version degree, xAm is the mean TAG conversion degree and n is the
was poured into a separation funnel in order to separate into the two number of experimental runs.
liquid phases. The crude biodiesel phase was then treated with Na2CO3 The pair sample t-test at the 0.05 level was employed to compare
and methanol (5% and 50% of biodiesel weight, respectively) while two means while all correlations were statistically tested by the
stirred at 65 °C for 4 h [44]. The biodiesel phase was separated from the Student’s t-test.
methanol phase in a separation funnel and washed with cold distilled
water (10% of biodiesel weight) while stirred (500 rpm) for 1 h. After 4. Results and discussion
separating water by centrifugation at 3500 rpm (average 700 × g) for
10 min, the washed biodiesel was dried with anhydrous sodium sulfate, 4.1. Analysis of WMSO transesterification over quicklime
which was separated by filtration.
The change of the TAG, DAG, MAG and FAME contents in the esters
phase of the reaction mixture with time is shown in Fig. 1. As it was
3.4. Analytical methods
expected, the FAME content varied sigmoidally. Initially, the rate of
FAME formation increased slowly, then accelerated and reached the
3.4.1. HPLC analysis of the ester/oil phase
maximum as the reaction approached the completion. On the other
The chemical composition of the ester phase of the reaction mixture
hand, as a consequence of FAME formation, the TAG content decreased
was determined by the HPLC method described in details elsewhere
during the reaction. The MAG and DAG contents passed through the
[42]. The experimental error of FAME content determination in the
maximum values but they were very low during the reaction.
replicated runs was ± 1.0%.

4.1.1. Effect of catalyst amount on the FAME formation

3.4.2. Characterization of WMSO biodiesel Fig. 2 shows the variation of FAME content with time during the
The physical and chemical properties of the purified biodiesel were WMSO transesterification catalyzed by quicklime at the methanol-to-
determined according to the appropriate standard methods, namely
density (EN ISO 3675:1988), kinematic viscosity (EN ISO 3104:2003),
iodine value (EN 14111:2003), acid value (EN 14104:2003), water
content (EN ISO 12937:2000), FAME content (EN 14103:2003), as well
as MAG; DAG and TAG contents (EN 14105:2003). All measurements
were performed in duplicate.

3.5. Calculations of model parameters

Values of the kinetic parameters of all models were calculated on

the basis of the experimental values of TAG conversion degree de-
termined with the progress of the transesterification reaction. The
parameters of the pseudo first-order models, Eqs. (3) and (5), were
calculated by the linear regression using the Polymath software. On the
other hand, the kinetic parameters of the model combining the chan-
ging mechanism with the pseudo first-order reaction rate law, the ap-
parent reaction rate constant, km , and the parameter defining the TAG
affinity for the catalyst active sites, K , were estimated from Eq. (11) by
the nonlinear regression using the Polymath software and minimizing
the objective function by the Levenberg–Marquardt algorithm. The
apparent reaction rate constant, kapp,2 , was calculated from Eq. (14) by
the linear regression using the Polymath software.
The same software and equations were used for calculation of the Fig. 1. The variation of the reaction mixture composition with the progress of the WMSO
predicted values of TAG conversion degree, xA , during the reaction. transesterification with methanol over quicklime (methanol-to-WMSO molar ratio: 6:1,
Then, concentrations of TAGs and FAMEs during the reaction were catalyst amount: 10% and temperature: 60 °C; FAME – ♢, MAG – •, DAG – ▴ and TAG –
■).
calculated from the predicted TAG conversion degree.

129
M.D. Kostić et al. Fuel 223 (2018) 125–139

Fig. 2. The variation of FAME content with the progress of WMSO transesterification Fig. 3. The variation of FAME content with the progress of WMSO transesterification
catalyzed by quicklime at the methanol-to-WMSO molar ratio of (a) 6:1, (b) 9:1 and (c) catalyzed by quicklime at the catalyst amount of (a) 2%, (b) 6%; (d) 10%, various the
12:1, various catalyst amounts (2% – ●, 6% – ▴ and 10% – ■) and the reaction tem- methanol-to-WMSO molar ratios (6:1 – ●, 9:1 – ▴ and 12:1 – ■) and the reaction tem-
perature of 60 °C. perature of 60 °C.

130
M.D. Kostić et al. Fuel 223 (2018) 125–139

WMSO molar ratios of 6:1, 9:1 and 12:1, various catalyst amounts (2%,
6% and 10%) and the reaction temperature of 60 °C. Generally, the
increase in the catalyst amount at a constant methanol-to-WMSO molar
ratio accelerated the reaction and shortened the time needed for com-
pleting the TAG conversion. However, Fig. 2c shows that the catalyst
amount larger than 6% has a slight influence on the overall process
rate, probably because of reducing the available catalytically active
surface [32]. The increase of FAME formation with increasing the cat-
alyst loading has already been reported [33,45–48]. However, reaching
the plateau [47,49–51] or even reducing the conversion above the
critical catalyst loading [52,53] was also observed. This was explained
by higher viscosity of the reaction mixture which lowered the mixing
efficiency [53] and favored the mass transfer limitation [51,54].

4.1.2. Effect of methanol-to-WMSO oil molar ratio on the FAME formation

The variation of FAME content during the WMSO transesterification
catalyzed by quicklime at the catalyst amount of 2%, 6% and 10%,
various methanol-to-WMSO molar ratios (6:1, 9:1 and 12:1) and the
reaction temperature of 60 °C is presented in Fig. 3. The impact of the
methanol-to-WMSO molar ratio on the FAME formation depended on
the catalyst amount. At the catalyst amounts of 2% and 6% (Fig. 3a and
b), the increase of the methanol-to-WMSO molar ratio from 6:1 to 12:1
accelerated the FAME formation, which was attributed to the effect of
the methanol excess on the reduction of the density and viscosity of the
reaction mixture. In turn, more efficient mixing accelerated the mass
transfer and improved the availability of the catalytically active centers
to the reacting molecules. Consequently, the FAME formation rate was
enhanced, thus shortening the initial lag period but the reaction was
completed almost in the same time. Slightly higher FAME content was
observed at the highest initial methanol-to-WMSO molar ratio, since the
reaction equilibrium was shifted towards the FAME formation. Similar
results were reported by Liu et al. [55] and Miladinović et al. [33],
respectively for the transesterification of soybean oil over calcium
methoxide and sunflower oil over quicklime. With the highest catalyst
loading (10% of the oil weight), the transesterification rate and FAME
content did not depend on the methanol-to-WMSO molar ratio (Fig. 3c).
This might be due to the combined effect of the reduced catalyst con-
centration in the reaction mixture and the more efficient mixing at
higher methanol-to-WMSO molar ratios, which opposed to each other
[33]. The effect of the methanol-to-oil molar ratio on FAME yield has
rather been investigated so far than its impact on the transesterification
rate. FAME yields increase with increasing the methanol-to-oil ratio
only up to a certain level in the range from 9:1 to 12:1 [51,53,56,57].

4.2. Kinetics of quicklime catalyzed sunflower oil methanolysis

4.2.1. Pseudo first-order law in heterogeneous and pseudo-homogeneous

regimes
Eqs. (3) and (5) can be used to estimate the volumetric TAG mass
transfer coefficient, kc a , and the apparent pseudo-first order reaction
rate constant, kapp,1, characterizing the mass transfer- and chemical
reaction-controlled regime, respectively by plotting −ln(1−xA) versus
time. As illustrated in Fig. 4, there is an excellent fit to the pseudo first-
order kinetic model for both regimes. Table 3 shows values of the rate
constant for both regimes obtained at reaction temperature of 60 °C and
various methanol-to-WMSO molar ratios and catalyst loadings.
Fig. 5 illustrates the dependence of the volumetric TAG mass
transfer coefficient divided by the initial methanol concentration,
kc a/ cB0 , and the apparent pseudo-first order reaction rate constant,
kapp,1, on the catalyst concentration at various initial methanol con-
Fig. 4. Plots of −ln(1−xA) versus time during the WMSO transesterification over quicklime
centrations. It is obvious from Fig. 5a that the volumetric TAG mass
at the methanol-to-WMSO molar ratio of (a) 6:1, (b) 9:1 and (c) 12:1 (catalyst amount,
transfer coefficient increases proportionally to the initial catalyst and
based on the oil weight, %: 2 – circle; 6 – triangle and 10 – square; irreversible pseu-
methanol concentrations (R2 = 0.985): do–first order reaction – straight lines; experimental data – solid symbols and sigmoidal
kc a = 0.00129·ccat ·cB0 (18) fit – open symbols).

This means that the volumetric TAG mass transfer coefficient, kc a , is

131
M.D. Kostić et al. Fuel 223 (2018) 125–139

Table 3
Values of the model parameters k c a and k app,1 at 60 °C.

Methanol-to-WMSO Catalyst, % kc a, R2 k app,1 , R2

molar ratio min−1 min−1

6:1 2 0.00140 0.977 0.1027 0.999

6 0.00513 0.990 0.1095 0.996
10 0.00751 0.916 0.1225 0.985

9:1 2 0.00260 0.941 0.1092 0.996

6 0.00619 0.993 0.1155 0.994
10 0.00955 0.898 0.1277 0.986

12:1 2 0.00460 0.842 0.1095 0.988

6 0.00732 0.800 0.1160 0.995
10 0.00988 0.906 0.1180 0.980

0.1145 ± 0.0077

dependent on both initial catalyst and methanol concentrations. On the

other hand, Fig. 5b shows the independence of the apparent pseudo-
first order reaction rate constant, kapp,1, on both initial catalyst and
methanol concentrations. Its mean value was determined to be
0.1145 ± 0.0077 min−1. The dependency of the volumetric TAG mass
transfer coefficient, kc a , and the independency of the apparent pseudo-
first order reaction rate constant, kapp,1, on the initial catalyst con-
centration have already been reported [32].
For the calculation of TAG conversion degree, the corresponding
pseudo first-order model was employed with the values of the volu-
metric TAG mass transfer coefficient, kc a , calculated by Eq. (18) and
the mean value of the apparent pseudo-first order reaction rate con-
stant, kapp,1. Hence, the values of the TAG conversion degree for the TAG
mass transfer and chemical reaction controlled regimes were calculated
by the following equations, respectively:
xA = 1−exp(−0.00129·ccat ·cB0·t ) (19)
and
xA = 1−exp(−0.1145·t −C1) (20)
Despite the excellent agreement with the experiment, the pseudo
first-order models suffer from a great drawback. Namely, they are valid
only in the corresponding regimes, which are not well-defined. Hence,
there is a middle region between the mass transfer-and chemical reac-
tion-controlled regimes where they are not applicable. In other words, Fig. 5. The dependence of a) k c a/ cB0 and b) k app,1 on the catalyst concentration at various
it is difficult to determine the end of the TAG mass transfer controlled initial methanol concentrations (methanol-to-WMSO molar ratio: 6:1 – •, 9:1 – ▴and 12:1
regime and the beginning of the chemical reaction controlled regime. – ■).

4.2.2. Changing mechanism combined with TAG mass transfer limitation

The calculated values of the kinetic parameters of the model com-
bining changing mechanism with TAG mass transfer limitation, km and
K (cR0 = 0 ), are given in Table 4. It may be noticed that km and K in-
creased with increasing both the catalyst loading and methanol-to-
WMSO molar ratio. To check if Eqs. (7) and (8) were valid, km/ cB0 and
K / cB0 are plotted against the catalyst concentration and the linear de-
pendences were obtained, as it can be seen in Fig. 6. The slopes of the
linear plots represented the reaction rate constant and the “pure” TAG
affinity for the catalyst active sites, i.e. k = 0.0312 L2/(mol2min) and
K′ = 0.424 L/mol, respectively. High values of the coefficients of de-
termination (R2 = 0.976 for both km/ cB0 and K / cB0 plots, respectively
with p < 0.001 in both cases) validated Eqs. (7) and (8). These values
of k and K ′ are smaller than those determined for the transesterification
of sunflower oil over quicklime, which are k = 0.045 L2/(mol2min) and
K′ = 0.813 L/mol [33]. It was shown the reaction rate constant k de-
pended on the type of the feedstock [41].
The values of the apparent reaction rate constant, kapp,2 , calculated
using the simpler linear equation based on the assumption of K ≫ cA ,
Eq. (14), are also given in Table 4. Its value appeared to be a constant
(mean kapp,2 = 0.0685 ± 0.0085 L/mol/min) for the ranges of the re-
action conditions applied in the present study. Since this constant re-
presents the ratio km/ K , its values can be compared with the corre-
sponding values calculated from the parameters km and K . The mean
km/ K -value of is 0.0722 ± 0.0044 L/mol/min, which is about 5%
higher than the mean kapp,2 -value. However, the pair sample t-test
showed that at the 0.05 level, these two means were not significantly
different to each other (t-statistic = −2.195, p = 0.059 > 0.050).
Hence, the simpler kinetic model, Eq. (14), can successfully replace the
more complex kinetic model, Eq. (10), for describing the kinetics of the
WMSO transesterification whereas the calculation of the model lumped
parameter, kapp,2 , is easier and more reliable.
The validity of the assumption of K ≫ cA is proved in Fig. 7, where
the TAG concentration and the parameter K are compared during the
transesterification reaction catalyzed by quicklime (2% and 10%). The
TAG concentration decreased rapidly in the medium and final phases of
the reaction, where the assumption of K ≫ cA was fulfilled.

132
M.D. Kostić et al. Fuel 223 (2018) 125–139

Table 4
Values of the model parameters km and K (cR0 = 0 ) calculated using Eq. (11).

Methanol-to-WMSO molar ratio Catalyst, km , K, R2 km/ K , k app,2 ,

% min−1 mol/L L/(mol min) L/(mol min)

6:1 2 0.071 1.03 0.957 0.0689 0.0578

6 0.109 1.50 0.963 0.0727 0.0631
10 0.177 2.50 0.973 0.0708 0.0788

9:1 2 0.070 1.00 0.979 0.0700 0.0660

6 0.105 1.40 0.962 0.0750 0.0615
10 0.225 3.12 0.997 0.0721 0.0797

12:1 2 0.072 1.12 0.947 0.0643 0.0647

6 0.171 2.21 0.978 0.0774 0.0657
10 0.301 3.83 0.991 0.0786 0.0796

0.0722 ± 0.0044 0.0685 ± 0.0072

4.3. Simulation of sunflower oil methanolysis

4.3.1. Simulation by the pseudo first-order reaction rate laws

The validity of the pseudo first-order kinetic model was checked by
comparing the calculated with experimental values of TAG conversion
degree with the progress of the transesterification reaction. In Fig. 8, a
good agreement between the predicted and actual values of TAG con-
version degree can be observed in both mass transfer and chemical
reaction controlled reaction regimes. The observed MRPD for TAG
conversion degree was only ± 3.0% (based on 147 data), confirming
the validity of the kinetic model. However, the pseudo first-order ki-
netic model was not applicable in the middle reaction period. Besides
that, the beginning and the end of this reaction period cannot easily be
recognized.

Fig. 6. Plots of km/cB0 (open symbols) and K/CB0 (black symbols) with the catalyst con-
centration at various methanol-to-WMSO molar ratios: 6:1 – circles, 9:1 – triangles and
12:1 – squares.
4.3.2. Simulation by the model combining the changing mechanism with the
TAG mass transfer limitation
In order to verify the proposed kinetic model combining the chan-
ging mechanism with the pseudo first-order reaction rate law with the
assumption of cR0 = 0 , the TAG conversion degree calculated on the
basis of this model was compared with the experimental data. Fig. 9
illustrates a good agreement between the model and the experiment. It
was found that the MRPD for TAG conversion degree was ± 13.0%
(based on 155 data), confirming the validity of the kinetic model.
A good agreement between the simplest model based on the com-
bination of the changing mechanism and the pseudo first-order reaction
rate law assuming both cR0 = 0 and K ≫ cA , Eq. (14), with the experi-
ments can be observed in Figs. 10 and 11 showing the variations of TAG
conversion degree and TAG and FAME concentrations during the
WMSO transesterification over quicklime, respectively. For the TAG
conversion degree, the MRPD of ± 16.1% (based on 155 data) proved
the capability of this simple model to describe the variation of xA during
the transesterification of WMSO with methanol over quicklime. Since
the inherent characteristic of this model is the validity in the middle
and later phases of the transesterification reaction, where K ≫ cA
(Fig. 7), this MRPD-value is acceptable. When the model was applied
only for the middle and final reaction phase (after 20 min), the MRPD-
value was ± 6.1% (based on 119 data). Hence, this simpler model, Eq.
(14), can replace successfully the more complex model, Eq. (11).

4.4. Statistical modeling and optimization of WMSO transesterification

The experimental matrix of the used 33 full factorial design with

Fig. 7. Comparison of the TAG concentration and the parameter K during the transes-
terification reaction (2%, 6:1: ● and 10%, 12:1: ■; model: solid lines; and K: dash line).
replication with actual and coded levels of the three factors along with
the achieved and predicted FAME contents is shown in Table 5. The
Kolmogorov–Smirnov normality test proved the analyzed data set was
significantly drawn from a normally distributed population at the 0.05
level (statistic = 0.128 < p = 0.314) (Fig. S1a, Supplementary mate-
rial). Besides that, this data set contained no outlier value (Fig. S1b,

133
M.D. Kostić et al.
Fig. 8. The comparison of the irreversible pseudo-first kinetic models (line) and the ex-
perimental data (symbols) during methanolysis at the methanol-to-oil molar ratio of a)
6:1, b) 9:1 and c) 12:1 (catalyst amount,%: 2 – ●, 6 – ▴ and 10 – ■).
Supplementary material).
The preliminary ANOVA was based on the full quadratic model, Eq.
(15). As it can be seen in Table S1 (Supplementary material), the
ANOVA results showed that at the 95% confidence level, the significant
terms were all three individual process factors (X1, X2 and X3), the in-
teractions of catalyst loading with methanol-to-WMSO molar ratio
(X1–X2) and reaction time (X2–X3), as well as the quadratic terms of
catalyst loading ( X22 ) while the interaction of methanol-to-WMSO molar
ratio with reaction time (X1–X3) and the quadratic terms of methanol-
to-WMSO molar ratio ( X12 ) and reaction time ( X32 ) had no statistically
significant influence on FAME content in the applied experimental
134
Fuel 223 (2018) 125–139
Fig. 9. The variation of TAG conversion degree with the progress of WMSO transester-
ification catalyzed by quicklime at the methanol-to-WMSO molar ratio of a) 6:1; b) 9:1
and c) 12:1, various catalyst amounts (2% – circles, 6% – triangles and 10% – squares)
and the reaction temperature of 60 °C (model – line).
region. The statistical criteria proved the goodness of fit and adequacy
of the full quadratic model (R2 = 0.947; Radj
2 2
= 0.936; Rpred = 0.917 ; and
MRPD = ± 12.9% based on 54 data). The lack of fit was insignificant
(p = 0.093), indicating that the quadratic model can be used for mod-
eling and optimization. Similar results were obtained for the transes-
terification of waste cooking oil over CaO, where all process factors
(catalyst amount, methanol-to-oil molar ratio and reaction time), their
squares and interaction had statistically significant impacts on the
biodiesel yield [58]. On the other hand, for the transesterification of
refined palm oil over CaO, the methanol-to-oil molar ratio and all two-
way interactions were statistically insignificant [59].
M.D. Kostić et al. Fuel 223 (2018) 125–139

Fig. 10. Comparison of the model combining the changing mechanism with the pseudo
first-order reaction rate law assuming K ≫ cA, Eq. (14), with the experiments carried out
at various catalyst loadings (% of oil): a) 2%, b) 6% and c) 10% and methanol-to-oil molar
ratios (model: line, and experiment: 6:1 – ●, 9:1 – ▴ and 12:1 – ■) at the reaction
temperature of 60 °C.
Fig. 11. Variations of TAG and FAME concentrations (black and open symbols, respec-
tively) during the WMSO transesterification over quicklime at the catalyst loadings (% of
oil): a) 2%, b) 6% and c) 10% and methanol-to-oil molar ratios (model: line, and ex-
periment: 6:1 – circles, 9:1 – triangles and 12:1 – squares) and the reaction temperature of
60 °C.

When the insignificant terms were eliminated, the simplified Y = 69.32 + 6.29·X1 + 22.96·X2 + 20.05·X3−6.67·X1 ·X2 −4.18·X2 ·X3
models, called here reduced quadratic equations, in terms of coded and −7.93·X22 (22)
actual factors were obtained:
The corresponding ANOVA results, based on the reduced quadratic
- Actual factors model, are presented in Table 6. The Fmodel- and p-values of 58.36
and < 0.0001, respectively meant the model was significant with only a
Y = −137.12 + 5.43·X1 + 20.87·X2 + 2.63·X3−0.56·X1 ·X2 −0.10·X2 ·X3
0.01% chance that this large Fmodel-value could occur due to noise.
−0.50·X22 (21) Besides that, the F-value of the lack of fit was insignificant (p-
value = 0.111 > 0.050), which was desirable, and hence, the reduced
- Coded factors quadratic model could be used to predict FAME content. Also, a high

135
M.D. Kostić et al. Fuel 223 (2018) 125–139

Table 5 cooking [58] and refined palm [59] oil over CaO. For the former case,
Experimental matrix of the 33 full factorial design with replication along with actual and the reaction time had the largest influence on biodiesel yield, then the
predicted WMSO content.a
methanol-to-oil molar ratio and finally the catalyst amount. Similarly,
Run Experimental matrix Response in the latter case, the reaction time showed the greatest impact on
biodiesel yield, followed by the catalyst loading and the methanol-to-oil
Uncoded factors Coded factors FAME content (%) molar ratio.
The relationship of FAME content with the methanol-to-WMSO
X1 X2 X3 X1 X2 X3 Actual, Actual, Predicted
series 1 series 2
molar ratio ( X1) and catalyst loading ( X2 ) at reaction time of 50 min is
graphically presented in Fig. 12 in the form of a response surface 3D
1 −1 −1 −1 6:1 2.0 30.0 7.4 7.3 1.2 plot along with the contour plot that resulted from the reduced quad-
2 0 −1 −1 9:1 2.0 30.0 15.2 12.7 14.2 ratic model. Generally, these plots visualize the effects of the process
3 1 −1 −1 12:1 2.0 30.0 28.0 34.4 27.2
4 −1 0 −1 6:1 6.0 30.0 38.5 36.3 43.0
factors and their interactions on FAME content and the optimal process
5 0 0 −1 9:1 6.0 30.0 48.0 43.0 49.3 conditions. In general, increasing the methanol-to-WMSO molar ratio,
6 1 0 −1 12:1 6.0 30.0 56.6 42.3 55.6 catalyst loading and reaction time led to an increase of FAME content.
7 −1 1 −1 6:1 10.0 30.0 74.7 54.0 68.9 For the selection of the optimal operating conditions using the de-
8 0 1 −1 9:1 10.0 30.0 78.1 70.1 68.5
veloped reduced quadratic model, the criterion of optimization was to
9 1 1 −1 12:1 10.0 30.0 74.1 52.5 68.1
10 −1 −1 0 6:1 2.0 40.0 19.8 20.4 25.5 get the maximum FAME content with the process factors constrained to
11 0 −1 0 9:1 2.0 40.0 30.2 34.3 38.4 the applied experimental region. According to the 3D plot (Fig. 12a),
12 1 −1 0 12:1 2.0 40.0 46.3 60.4 51.4 the FAME content higher than 96.5% can be obtained in 50 min in the
13 −1 0 0 6:1 6.0 40.0 73.2 64.6 63.0 entire applied molar ratio (from 6:1 to 12:1) if the catalyst loading was
14 0 0 0 9:1 6.0 40.0 71.6 71.8 69.3
15 1 0 0 12:1 6.0 40.0 82.2 73.4 75.6
appropriately selected (approximately larger than 9%). For shorter
16 −1 1 0 6:1 10.0 40.0 90.9 83.2 84.7 times, much lower FAME content could be reached (for instance < 83%
17 0 1 0 9:1 10.0 40.0 96.5 92.4 84.4 in 40 min). The used software on the basis of the reduced quadratic
18 1 1 0 12:1 10.0 40.0 95.5 82.9 84.0 model suggested a number of combinations, as it can be seen in Table 7.
19 −1 −1 1 6:1 2.0 50.0 49.7 50.7 49.7
According to the reduced quadratic model, complete conversion can be
20 0 −1 1 9:1 2.0 50.0 63.2 56.2 62.7
21 1 −1 1 12:1 2.0 50.0 75.0 80.5 75.6 achieved in 50 min if the catalyst loading is 9.8% or larger and the
22 −1 0 1 6:1 6.0 50.0 90.9 88.3 83.1 methanol-to-WMSO molar ratio is in the range between 6.1:1 and
23 0 0 1 9:1 6.0 50.0 92.3 89.5 89.4 11.6:1. The actual FAME contents obtained in 50 min at the catalyst
24 1 0 1 12:1 6.0 50.0 94.4 90.8 95.7 loading of 10% and the methanol-to-WMSO molar ratios of 6:1 and 12:1
25 −1 1 1 6:1 10.0 50.0 94.9 91.0 100.6
were 93.0% and 96.5%, indicating overestimation of FAME content by
26 0 1 1 9:1 10.0 50.0 98.4 96.3 100.2
27 1 1 1 12:1 10.0 50.0 98.4 94.5 99.8 the model. In order to decrease the cost of the excess methanol se-
paration from the final reaction mixture, the methanol-to-WMSO molar
a
The predicted FAME content was calculated using the reduced quadratic equation. ratio of 6:1 was selected as the optimum one.

R2-value (0.944) and low MRPD-value ( ± 12% on the basis of 54 data)

proved the adequacy of the reduced quadratic model. Moreover, the 4.5. Comparison of WMSO transesterification over quicklime and KOH
2 2
Rpred - and Radj -values (0.924 and 0.937, respectively) were close to each
other as normally expected [60], indicating a good prediction of FAME Alkali hydroxides are most frequently used as catalysts for com-
by Eqs. (21) and (22). mercial transesterification of vegetable oils. Therefore, it is encouraging
As it can be seen in Eq. (22), the linear regression coefficients were to compare the efficiencies of the transesterification of WMSO catalyzed
positive, indicating a positive influence of all three process factors on with quicklime and the traditional ones. Fig. 13 shows the progress of
FAME content. The values of the linear regression coefficients of Eq. the transesterification of WMSO (this work) and sunflower oil [61,62]
(22) and the F-values of the individual process factors (Table 6) in- in the presence of KOH (1%) or quicklime (10%) at the methanol-to-oil
dicated that the influence of catalyst loading and reaction time was molar ratio of 6:1 and the reaction temperature of 60 °C or 65 °C. It is
similar and larger than the effect of methanol-to-WMSO molar ratio obvious that the homogeneous reactions are faster than the hetero-
(X2 ≈ X3 > X1). Also, it was observed that higher FAME contents were geneous ones due to the more intensive mass transfer limitations in the
achieved at higher catalyst loading and methanol-to-WMSO molar ratio liquid-liquid-solid systems. Besides the mass transfer limitation be-
and for longer reaction time. Different impact of the process factors on tween the two immiscible liquids, both intraparticle and external li-
the FAME yield was observed for the transesterification of waste quid-solid mass transfer limitations might control the overall rate of the
heterogeneous catalysis [32]. Fig. 13 clearly shows the negligible mass
Table 6 transfer control region in the beginning of the KOH-catalyzed reaction
ANOVA results: reduced quadratic model. while it is easily observed in the case of quicklime. This fact has already
Source Sum of squares df Mean square F-value p-value
been shown in many studies that compare homogeneous and hetero-
geneous catalysts [31]. The obvious slower reaction rate and hence
Model 37121.0 6 6186.8 131.7 < 0.0001 longer reaction time in the presence of quicklime is compensated by
X1 1423.8 1 1423.8 30.3 < 0.0001 simpler, cheaper and more environmentally-friendly overall process.
X2 18984.2 1 18984.2 404.0 < 0.0001
X3 14472.1 1 14472.1 308.0 < 0.0001
On the other hand, the WMSO was transesterified faster and with a
X1 X2 1068.0 1 1068.0 22.7 < 0.0001 shorter initial lag period than sunflower oil in the presence of both KOH
X2 X3 419.2 1 419.2 8.9 0.004 and quicklime, which can be attributed to the differences in their
X22 753.7 1 753.7 16.0 < 0.001 compositions. For instance, the total content of unsaturated fatty acids
Residual 2208.6 47 47.0 of sunflower oil and WMSO were 88.6% [63] and 98% [13], respec-
Lack of Fit 1215.8 20 60.8 1.7 0.111
tively. The higher equilibrium WMSO conversion degree in shorter re-
Pure Error 992.8 27 36.8
Corrected Total 39329.6 53 action time, compared to sunflower oil, resulted from its higher total
content of unsaturated fatty acids.
2 2
R2 = 0.944; Radj = 0.937; Rpred = 0.924; C.V. = 10.7%; and MRPD = ± 12% (based on
54 data).

136
M.D. Kostić et al. Fuel 223 (2018) 125–139

Fig. 12. Response surface (a) and contour (b) plots for the reduced quadratic model (reaction time: 50 min; experimental points: ●).

Table 7 4.6. Biodiesel quality

Optimal reaction conditions on the basis of the reduced quadratic model for the WMSO
transesterification over quicklime. The characteristics of the purified biodiesel obtained from WMSO
Run X1 X2 X3 FAME, %
by the transesterification catalyzed by quicklime and KOH are pre-
sented in Table 8. As it can be seen, both biodiesel products satisfied the
mol/mol % of oil min Predicted Actual biodiesel standard specifications. For comparison, the reported prop-
erties of the purified ester phases obtained by the alkali-catalyzed
1 6.1 9.8 50 100 93.0a
2 6.4 9.9 50 100
transesterification of WMSO with methanol in the presence of KOH [8],
3 6.8 9.8 50 100 NaOH [19,28] and potassium methoxide [15] are also given in Table 8.
4 7.4 9.8 50 100 Despite a high removal efficiency (97%) of the group II metals
5 8.1 9.8 50 100 (Ca + Mg) from crude biodiesel obtained in the quicklime-catalyzed
6 9.0 9.9 50 100
WMSO transesterification process, its content in the purified biodiesel
7 10.0 9.9 50 100
8 10.7 9.8 50 100 was above the specified limit. Therefore, the purification process should
9 11.5 9.4 50 100 be improved in order to further reduce the present calcium and mag-
10 11.6 9.0 50 100 96.5b nesium. The content of the group I metals (Na + K) in the purified
a
biodiesel obtained in the KOH-catalyzed WMSO transesterification
X1 = 6, X2 = 10 and X3 = 50.
b process satisfied the standard specification. The other properties of both
X1 = 12, X2 = 10 and X3 = 50.
purified biodiesels fulfilled the standard limits and were similar to those
reported by other research groups.

5. Conclusion

Biodiesel production in the future will depend on low-cost, non-

edible and renewable feedstocks and low-cost, active, stable and solid
catalysts that will be used in novel more intensive technologies. Having
the required properties, WMSO and quicklime are shown in the present
study as good candidates for the biodiesel production. WMSO was
transesterified even faster and with a shorter initial lag period than
sunflower oil in the presence of both quicklime and KOH, due to higher
total content of unsaturated fatty acids. Under the optimum reaction
conditions in a batch stirred reactor, the complete conversion can be
achieved in 50 min if the catalyst loading is 9.8% or larger and the
methanol-to-WMSO molar ratio is in the range between 6.1:1 and
11.6:1.
The kinetics of the WMSO transesterification was described by the
two known models: a) two pseudo-first order rate laws with respect to
TAGs for mass transfer- and chemical reaction-controlled regimes and
b) the changing mechanism combined with the first-order rate law with
respect to TAGs and FAMEs, respectively. The second model was sim-
Fig. 13. Effect of the catalyst type on the transesterification of various vegetable oils like
plified to enable more reliable calculation of the kinetic model para-
WMSO (circles) and sunflower oil (triangles) up in the presence of quicklime [62] or KOH
meter.
[61] (quicklime: black symbols and KOH: open symbols; catalyst loading, of the oil
weight: 1% KOH and 10% quicklime; the methanol-to-oil molar ratio: 6:1; and the re-
action temperature: 60 °C or 65 °C applied by Vicente et al. [61]).

137
M.D. Kostić et al. Fuel 223 (2018) 125–139

Table 8
Properties of biodiesel produced from WMSO.

Property, unit Reactor, catalyst, methanol-to-oil molar ratio, temperature and purification methoda EN14214 limits,
min/max
Batch, CaO (10%), Batch, KOH, (1%), Batch, KOH, Batch, NaOH Batch, NaOH Batch, CH3OK (0.2:1
12:1, 60 °C, AR 6:1, 60 °C, WW (1%), 6:1, (0.5%), 6:1, (0.72%), 6:1, catalyst-to-oil molar
60 °C, WW 65 °C, WW 70 °C, WW ratio), 6:1, 60 °C, WW

Density at 15 °C, kg/m3 881.1 ± 0.8 880.1 ± 0.7 900 834.3 899 878 860/900
Viscosity at 40 °C, 4.15 4.13 4.2 5.45 6.72 5.67 3.50/5.00
mm2/s
Acid value, mg KOH/g 0.44 ± 0.03 0.47 ± 0.04 0.4 0.8 – 0 0.50 max
Iodine value, g I2/100 g 102.9 ± 0.6 104.7 ± 1.6 – – – 102.3 120 max
Water content, mg/kg 235 ± 7 217 ± 4 234 – – 223 500 max
FAME content, % 98.5 98.7 99.8 – 82 > 98 96.5 min
MAG content, % 0.5 0.4 0 – – 0.15 0.80 max
DAG content, % 0.1 0.1 0 – – 0.05 0.20 max
TAG content, % 0.2 0.2 0 – – 0.02 0.20 max
Group I metals – 3.7 – – – – 5.0 max
(Na + K), mg/kg
Group II metals 15.5 – – – – – 5.0 max
(Ca + Mg), mg/kg
Reference This work [8] [19] [28] [15]

a
AR – Alba-Rubio et al. [44], WW – water washing.

Acknowledgment
This work has been funded by the Ministry of Education, Science
and Technological Development of the Republic of Serbia, Serbia
(Project III 45001).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.fuel.2018.03.023.
References
[1] Živković SB, Veljković MV, Banković-Ilić IB, Krstić IM, Konstantinović SS, Ilić SB,
et al. Technological, technical, economic, environmental, social, human health risk,
toxicological and policy considerations of biodiesel production and use. Renew Sust
Energy Rev 2017;79:222–47.
[2] Banković-Ilić IB, Stamenković OS, Veljković VB. Biodiesel production from non-
edible plant oils. Renew Sust Energy Rev 2012;16:3621–47.
[3] Moser BR, Williams A, Haas MJ, McCormick RL. Exhaust emissions and fuel prop-
erties of partially hydrogenated soybean oil methyl esters blended with ultra low
sulfur diesel fuel. Fuel Process Technol 2009;90:1122–8.
[4] Krstić Đ, Ćupina B, Antanasović S, Erić P, Čabilovski R, Manojlović M, Mikić A.
Potential of white mustard (Sinapis alba L. subsp. alba) as a green manure crop.
Crucif Newslett 2010;29:12–3.
[5] Balke DT, Diosady LL. Rapid aqueous extraction of mucilage from whole white
mustard seed. Food Res Int 2000;33:347–56.
[6] Yaniv Z, Schafferman D, Elber Y, Ben-Moshe E, Zur M. Evaluation of Sinapis alba,
native to Israel, as a rich source of erucic acid in seed oil. Ind Crop Prod
1994;2:137–42.
[7] Falasca LS, Ulberich A. Argentina’s semiarid lands aptitude to cultivate non-tradi-
tional species for biodiesel production. In: Marchetti JM, Fang Z, editors. Biodiesel:
blends, properties and applications. New York, USA: Nova Science Publishers Inc;
2011. p. 123–50.
[8] Issariyakul T, Dalai AK. Evaluating esters derived from mustard oil (Sinapis alba) as
potential diesel additives. J Am Oil Chem Soc 2011;88:391–402.
[9] Peng C, Zhang T, Zhao G, Wang S. Analysis on fat-soluble components of Sinapis
semina from different habitats by GC-MS. J Pharm Anal 2013;3:402–7.
[10] Issariyakul T, Dalai AK. Biodiesel from Mustard Oil. In: Thiyam-Holländer U, Eskin
NAM, Mattäus B, editors. Canola and rapeseed: production, processing, food
quality, and nutrition. Boca Roca, London, New York: CRC Press; 2012. p. 217–44.
[11] Sanjid A, Masjuki HH, Kalam MA, Abedin MJ, Ashrafur Rahman SM. Experimental
investigation of mustard biodiesel blend properties, performance, exhaust emission
and noise in an unmodified diesel engine. APCBEE Proc 2014;10:149–53.
[12] Peng C, Zhao S-Q, Zhang J, Huang G-Y, Chen L-Y, Zhao F-Y. Chemical composition,
antimicrobial property and microencapsulation of Mustard (Sinapis alba) seed es-
sential oil by complex coacervation. Food Chem 2014;165:560–8.
[13] Stamenković OS, Djalović IG, Kostić MD, Mitrović PM, Veljković VB. Extraction of
oil from white mustard (Sinapis alba L.) seeds: optimization and kinetic modeling.
Ind Crop Prod 2017. [submitted].
[14] Thacker PA, Petri D. The effects of canola or mustard biodiesel press cake on nu-
trient digestibility and performance of broiler chickens. Asian-Aust J Anim Sci
2009;22:531–9.
[15] Ciubota-Rosie C, Macoveanu M, Fernández CM, Ramos MJ, Pérez A, Moreno A.
Sinapis alba seed as a prospective biodiesel source. Biomass Bioenergy
2013;51:83–90.
[16] Sáez-Bastante J, Fernández-García P, Saavedra M, López-Bellido L, Dorado MP,
Pinzi S. Evaluation of Sinapis alba as feedstock for biodiesel production in
Mediterranean climate. Fuel 2016;184:656–64.
[17] Peterson CL, Thompson J. Biodiesel from mustard oil. National Institute for
Advanced Transportation Technology, University of Idaho; 2005. [Final Report No.
05-06].
[18] Tyson KS, Brown J, Moora M. Industrial mustard crops for biodiesel and biopesti-
cides. Report. Golden, CO: National Renewable Energy Laboratory; 2000. http://
infohouse.p2ric.org/ref/35/34337.pdf (assessed 19.02.2017).
[19] Sultana S, Khalid A, Ahmad M, Zuhairi AA, Teong LK, Zafar M, et al. The produc-
tion, optimization, and characterization of biodiesel from a novel source: Sinapis
alba L. Int J Green Energy 2014;11:280–91.
[20] Tabtabaei S, Boocock DGB, Diosady LL. Biodiesel feedstock from emulsions pro-
duced by aqueous processing of yellow mustard. J Am Oil Chem Soc
2014;91:1269–82.
[21] Tabtabaei S, Boocock DGB, Diosady LL. Biodiesel production from mustard emul-
sion by a combined destabilization/adsorption process. J Am Oil Chem Soc
2015;92:1205–17.
[22] Issariyakul T, Dalai AK. Comparative kinetics of transesterification for biodiesel
production from palm oil and mustard oil. Can J Chem Eng 2012;90:342–50.
[23] Alam MM, Rahman KA. Biodiesel from mustard oil: a sustainable engine fuel sub-
stitute for Bangladesh. Int J Renew Energ Dev 2013;2:141–9.
[24] Oshodi OA, Chukwuneke CE, Linus O. The viscometric analysis of biodiesel from
mustard and coconut oil. Eur Chem Bull 2014;3:946–8.
[25] Nie J, Wang S, Emami S, Falk K, Shen J, Reaney MJT. Unusually low pour point of
fatty acid methyl esters with low saturated fatty acid content. Eur J Lipid Sci
Technol 2016;118:1486–94.
[26] Sarala R, Rajendran M, Devadasan SR. Emission characteristics of mustard oil
methyl ester (MOME) – diesel fuel blends on a C.I Engine. Int J Green Chem
Bioprocess 2012;2:6–10.
[27] Ahmad M, Ajub Khan M, Zafar M, Hasan A, Ahmad Z, Akhter G, et al. Base cata-
lyzed transesterification of Brassica alba oil and its association with mineralogy to
environment friendly biodiesel. Asian J Chem 2008;20:6402–10.
[28] Ahmad M, Zafar M, Rashid S, Sultana S, Sadia H, Ajab Khan M. Production of
methyl ester (biodiesel) from four plant species of Brassicaceae: optimization of the
transesterification process. Int J Green Energy 2013;10:362–9.
[29] Banković-Ilić IB, Miladinović MR, Stamenković OS, Veljković VB. Application of
nano CaO-based catalysts in biodiesel synthesis. Renew Sust Energy Rev
2017;57:746–60.
[30] Kesić Ž, Lukić I, Zdujić M, Mojović L, Skala D. Calcium oxide based catalysts for
biodiesel production: a review. Chem Ind Chem Eng Q 2016;22:391–408.
[31] Marinković DM, Stanković MV, Veličković AV, Avramović JM, Miladinović MR,
Stamenković OS, et al. Calcium oxide as a promising heterogeneous catalyst for
biodiesel production: current state and perspectives. Renew Sust Energy Rev
2016;56:1387–408.
[32] Veljković VB, Stamenković OS, Todorović ZB, Lazić ML, Skala DU. Kinetics of
sunflower oil methanolysis catalyzed by calcium oxide. Fuel 2009;88:1554–62.
[33] Miladinović MR, Krstić JB, Tasić MB, Stamenković OS, Veljković VB. A kinetic study
of quicklime-catalyzed sunflower oil methanolysis. Chem Eng Res Des
2014;92:1740–52.
[34] Granados ML, Poves MDZ, Alonso DM, Mariscal R, Galisteo FC, Tost RM, et al.
Biodiesel from sunflower oil by using activated calcium oxide. Appl Catal B:

138
M.D. Kostić et al.
Environ 2007;73:317–26.
[35] Eze VC, Phan AN, Harvey AP. A more robust model of the biodiesel reaction, al-
lowing identification of process conditions for significantly enhanced rate and
water tolerance. Bioresour Technol 2014;156:222–31.
[36] Stamenković OS, Veljković VB, Todorović ZB, Lazić ML, Banković-Ilić IB, Skala DU.
Modeling the kinetics of calcium hydroxide catalyzed methanolysis of sunflower oil.
Bioresour Technol 2010;101:4423–30.
[37] Deshmane VG, Adewuyi YG. Synthesis and kinetics of biodiesel formation via cal-
cium methoxide base catalyzed transesterification reaction in the absence and
presence of ultrasound. Fuel 2013;107:474–82.
[38] Lukić I, Kesić Ž, Maksimović S, Zdujić M, Liu H, Krstić J, et al. Kinetics of sunflower
and used vegetable oil methanolysis catalyzed by CaO·ZnO. Fuel 2013;113:367–78.
[39] Lukić I, Kesić Ž, Maksimović S, Zdujić M, Krstić J, Skala D. Kinetics of hetero-
geneous methanolysis of sunflower oil with CaO ZnO catalyst: influence of different
hydrodynamic conditions. Chem Ind Chem Eng Q 2014;20:425–39.
[40] Kostić MD, Bazargan A, Stamenković OS, Veljković VB, McKay G. Optimization and
kinetics of sunflower oil methanolysis catalyzed by calcium oxide-based catalyst
derived from palm kernel shell biochar. Fuel 2016;163:304–13.
[41] Stojković IJ, Miladinović MR, Stamenković OS, Banković-Ilić IB, Povrenović DS,
Veljković VB. Biodiesel production by methanolysis of waste lard from piglet
roasting over quicklime. Fuel 2016;182:454–66.
[42] Veličković AV, Avramović JM, Stamenković OS, Veljković VB. Kinetics of the
sunflower oil ethanolysis using CaO as catalyst. Chem Ind Chem Eng Q
2016;22:409–18.
[43] Miladinović MR, Tasić MB, Stamenković OS, Veljković VB, Skala DU. Further study
on kinetic modeling of sunflower oil methanolysis catalyzed by calcium-based
catalysts. Chem Ind Chem Eng Q 2016;22:137–44.
[44] Alba-Rubio AC, Alonso Castillo ML, Albuquerque MCG, Mariscal R, Cavalcante Jr
CL, López Granados M. A new and efficient procedure for removing calcium soaps
in biodiesel obtained using CaO as a heterogeneous catalyst. Fuel 2012;95:464–70.
[45] Dias JM, Alvim-Ferraz MCM, Almeida MF, Díaz JDM, Polo MS, Utrilla JR. Selection
of heterogeneous catalysts for biodiesel production from animal fat. Fuel
2012;94:418–25.
[46] Liu H, Su L, Shao Y, Zou L. Biodiesel production catalyzed by cinder supported
CaO/KF particle catalyst. Fuel 2012;97:651–7.
[47] Wang J-X, Chen K-T, Wen B-Z, Liao Y-HB, Chen C-C. Transesterification of soybean
oil to biodiesel using cement as a solid base catalyst. J Taiwan Inst Chem Eng
2012;43:215–9.
[48] Wang B, Li S, Tian S, Feng R, Meng Y. A new solid base catalyst for the transes-
terification of rapeseed oil to biodiesel with methanol. Fuel 2013;104:698–703.
139
Fuel 223 (2018) 125–139
[49] Teng G, Gao L, Xiao G, Liu H. Transesterification of soybean oil to biodiesel over
heterogeneous solid base catalyst. Energ Fuel 2009;23:4630–4.
[50] Wei Z, Xu C, Li B. Application of waste eggshell as low-cost solid catalyst for bio-
diesel production. Bioresour Technol 2009;100:2883–5.
[51] Wu H, Zhang J, Wei Q, Zheng J, Zhang J. Transesterification of soybean oil to
biodiesel using zeolite supported CaO as strong, base catalysts. Fuel Process Technol
2013;109:13–8.
[52] Noiroj K, Intarapong P, Luengnaruemitchai A, Jai-In S. A comparative study of
KOH/Al2O3 and KOH/NaY catalysts for biodiesel production via transesterification
from palm oil. Renew Energy 2009;34:1145–50.
[53] Yang Z, Xie W. Soybean oil transesterification over zinc oxide modified with alkali
earth metals. Fuel Process Technol 2007;88:631–8.
[54] Yan S, Lu H, Liang B. Supported CaO catalysts used in the transesterification of
rapeseed oil for the purpose of biodiesel production. Energ Fuel 2008;22:646–51.
[55] Liu X, Piao X, Wang Y, Zhu S, He H. Calcium methoxide as a solid base catalyst for
the transesterification of soybean oil to biodiesel with methanol. Fuel
2008;87:1076–82.
[56] Sánchez-Cantú M, Pérez-Díaz LM, Pala-Rosas I, Cadena-Torres E, Juárez-Amador L,
Rubio-Rosas E, et al. Hydrated lime as an effective heterogeneous catalyst for the
transesterification of castor oil and methanol. Fuel 2013;110:54–62.
[57] Viriya-Empikul N, Krasae P, Puttasawat B, Yoosuk B, Chollacoop N, Faungnawakij
K. Waste shells of mollusk and egg as biodiesel production catalysts. Bioresour
Technol 2010;101:3765–7.
[58] Aworanti OA, Agarry SE, Ajani AO. Statistical optimization of process variables for
biodiesel production from waste cooking oil using heterogeneous base catalyst. Brit
Biotechnol J 2013;3:116–32.
[59] Suwanthai W, Punsuvon V. Optimization of transesterification reaction for biodiesel
production from refined palm oil using calcined quick lime catalyst. Asian J Chem
2016;28:423–8.
[60] Anderson MJ. Statistics Made Easy_Blog, June 2010; 10(6) < http://www.statease.
com/news/faqalert10-06.html > [accessed 15.05.2017].
[61] Vicente G, Martınez M, Aracil J, Esteban A. Integrated biodiesel production: a
comparison of different homogeneous catalysts systems. Bioresour Technol
2004;92:297–305.
[62] Miladinović MR, Tasić MB, Stamenković OS, Veljković VB, Skala DU. Further study
on kinetic modeling of sunflower oil methanolysis catalyzed by calcium-based
catalysts. Chem Ind Chem Eng Q 2016;22:137–44.
[63] Avramović JM, Stamenković OS, Todorović ZB, Lazić ML, Veljković VB. The opti-
mization of the ultrasound-assisted base-catalyzed sunflower oil methanolysis by a
full factorial design. Fuel Process Technol 2010;91:1551–7.