Chaudhary 2016
Chaudhary 2016
Chaudhary 2016
To cite this article: Payal Chaudhary, Ayushi Verma, Surendra Kumar & V.K. Gupta (2016):
Experimental design and optimization of castor oil transesterification process by response
surface methodology, Biofuels, DOI: 10.1080/17597269.2016.1249737
Article views: 2
Table 1. Some important studies on the optimization of different reaction conditions on % FAME yield.
Temperature Catalyst % (w/ Reaction time Agitator speed % FAME
S. no. Source Catalyst ( C) w) Molar ratio (min) (rpm) yield Ref.
1 Castor oil KOH 55–65 1 6:1 120–480 – 73.62 [8]
2 Castor oil KOH, NaOH, 25–65 0.5–1.5 5:1–7:1 10 600 99.81 [10]
CH3ONa, CH3OK
3 Castor oil KOH 24.8–75.2 0.16–1.84 2.45:1–12.55:1 40 300 92 [11]
4 Castor oil KOH 45–65 0.5–1.5 100:1–350:1 180 100–600 98.6 [12]
5 Rapeseed oil KOH 25–65 0.4–1.6 4:1–8:1 45–85 400 84.05 [13]
6 Moringa olifera oil KOH 25–65 0.25–1.25 3:1–12:1 20–90 600 94.30 [14]
8 Jatropha oil H2SO4 60 1.36–1.54 0.24:1–0.36:1 42.2–77.8 – >99.0 [15]
formation of methyl esters from castor oil using an acid and methanol as an alcohol for the transesterification
catalyst, concentrated H2SO4, a basic catalyst, KOH, and reaction of castor oil, since this combination has not
12-tungstophosphoric acid (H3PW12O40, HPW) as the been reported on in the literature yet.
homogeneous catalyst. The potassium-doped salts of In recent years, artificial neural networks (ANNs)
HPW proved to be the best catalyst for the formation have been applied to a wide range of applications
of methyl esters. such as data prediction, fault detection, process
Cavalcante et al. [19] studied the transesterification control and data rectification. Chakraborty and Sahu
of castor oil with ethanol using a central composite [9] evaluated the intensification of biodiesel produc-
rotatable design and they determined the optimum tion from waste goat tallow using RSM and ANN
reaction conditions as an oil:ethanol molar ratio of with appreciable modeling efficiency. Verma and
1:11, catalyst amount of 1.75% KOH, and reaction time Sharma [23] compared the process yield of karanja
of 90 min, and obtained 86% of biodiesel yield. Jeong oil with methanol and ethanol and obtained a total
et al. [11] used the response surface methodology biodiesel yield 91.05% with methanol and 77.40%
(RSM) for the optimization of biodiesel production with ethanol. Saydut et al. [24] studied the optimi-
from castor oil in the presence of CH3OK as catalyst. zation process for two oils, hazelnut oil and sun-
They obtained a 92% biodiesel yield by optimizing the flower oil, and their hybrid feedstock oil and
reaction conditions with a castor oil:methanol molar obtained total biodiesel yields of 97.50, 97.30 and
ratio of 1:8.24, catalyst amount of 1.45% CH3OK, reac- 97.90%, respectively.
tion time of 120 min, and reaction temperature of In the present study, a mathematical model was
35 C. Silva et al. [20] optimized the alkali ethanolysis of built for transesterification of castor oil in the presence
castor oil by using factorial design, and the optimized of an acid catalyst using RSM which comprises a central
reaction conditions were ethanol:castor oil molar ratio composite design (CCD) to describe the effects and the
of 16:1, catalyst amount of 1.0% C2H5ONa, reaction relationships between the process variables, methanol:
time of 30 min, and reaction temperature of 30 C, and oil molar ratio, catalyst amount and reaction tempera-
the total biodiesel yield obtained was 93%. Naik et al. ture, on the conversion of fatty acids, and thus to
[21] discussed the mechanism of the two-step process determine the operating conditions that maximize the
adopted for the production of biodiesel from Karanja % FAME yield.
oil with high free fatty acid (FFA) content. In this type
of reaction, the first step consists of the reaction of
fatty acids with an acid catalyst for the esterification Materials and methods
reaction to reduce the FFA content, and after this step Chemicals
the base catalyst was used for the transesterification
reaction. This dual-step process successfully resulted in Pharmaceutical-grade castor oil was purchased from
a 96.6–97% yield of biodiesel from karanja oil. Liu et al. the local market. All chemicals – H2SO4 (97%), anhy-
[22] investigated that acid catalysts have an important drous methanol, Na2SO4 – were of analytical grade and
advantage as they can simultaneously catalyze both were purchased from Merck India and used without
esterification and transesterification of triglycerides, further purification. Methyl heptadecanoate was used
and they can produce biodiesel from low-cost lipid as an internal standard. Pure standard of methyl esters
feedstock materials. Castor oil has a high FFA value was purchased from Sigma Aldrich, USA.
and the base catalyst could not be used directly with
the high amount of FFA, as the reaction of saponifica-
Experimental setup
tion will take place and soap will form as product.
When the acid catalyst is used for the esterification A bioreactor (Applikon, Schiedam, The Netherlands,
reaction it will not be affected by the presence of FFA stirred type, capacity: 3 L, EZ-control system) was used
and esterification as well as transesterification reac- for conducting the reaction. This reactor was a 3 L dou-
tions will proceed simultaneously. Thus the catalyst- ble-jacketed borosilicate glass reactor having 2 L work-
concentrated H2SO4 has been selected as the catalyst ing volume. The temperature was regularly monitored
BIOFUELS 3
by display of the system. The reflux system was used Analysis of all samples for methyl ester formation
to avoid the vaporization of methanol from a reac- (FAME content) was carried out using a gas chromato-
tion system. The temperature of the reaction system graph (GC, Nucon Gas Chromatograph, 5765, INDIA),
was controlled by the circulation of water through equipped with a flame ionization detector (FID) and a
outer wall of a reactor vessel. Mechanical stirring capillary column with dimension of 0.55 mm inner
was used for the proper mixing of the reaction mix- diameter £ 10 m length £ 0.50 m thickness. Nitrogen
ture in a reactor. The overall transesterification reac- was used as the carrier gas. The column temperature
tion mechanism of triglycerides with acid catalyst was kept at 170 C for 1 min, heated at 10 C/min up to
can be summarized as below: 240 C and then kept constant. The temperature of the
R3 R3
O O O O
+
O O H O OH+
(1)
R1 R1
O O
R2 O R2 O
R3 R3
H
O O O O H
O +
O OH+ O O
(2) + OR
R
R1 R1
O O
R2 O R2 O
R3
R3
O O
H
O O O + H OH + H
O O
(3) R R O
+ R
R1 O
O
O
R2 O
R2 O
R3 R3
O O O O
OH OH O
(4) + H R
R +
O + H
+ R
R +
O
O O
O
R2 O R2 O
Figure 1. Reaction mechanism of acid catalyzed transesterification of triglycerides with alcohol. (1) Protonation of the carbonyl
group by the acid catalyst; (2) nucleophilic attack of the alcohol, forming a tetrahedral intermediate, (3 and 4) proton migration
and the breakdown of the intermediate. R1, R2 and R3 refers to carbon chain of the fatty acid whereas R refers to alkyl group of
the alcohol [25,26].
4 P. CHAUDHARY ET AL.
injector and detector were set at 220 and 240 C, Table 2. Independent variables and levels used for CCD.
respectively. Methyl heptadecanoate was used as the Range and
S. no. Variables Coded symbols levels (¡1, 0, C1)
internal standard for GC. The analysis was done by 1 Methanol/oil molar ratio x1 6, 15.5, 25
injecting a 1-mL sample into the column. Quantitative 2 Catalyst amount (vol. %) x2 1, 2, 3
analysis of % methyl esters (ME) was done using Euro- 3 Temperature ( C) x3 40, 50, 60
independent variable factors was developed. Once all Table 4. Observed responses and predicted values.
the experiments were performed, the response vari- Run no. Actual response Predicted response Residual values
1 55.27 52.83 2.44
able (% FAME yield) was fitted in a second-order model 2 51.25 51.67 ¡0.42
to find out the correlation between the response vari- 3 51.54 51.67 ¡0.13
able and the independent variables. The general form 4 47.36 44.92 2.44
5 27.92 27.50 0.42
of the second-degree polynomial Equation (2) is as fol- 6 52.15 51.67 0.48
lows: 7 51.92 51.67 0.25
8 37.71 38.45 ¡0.74
9 48.29 49.03 ¡0.74
X
k X
k k X
X k 10 60.18 62.58 ¡2.40
Y D bo C bi xi C bij xi 2 C bij xi xj C e 11 31.45 32.19 ¡0.74
i D 1 i D 1 ii > 1 j 12 67.89 70.29 ¡2.40
13 51.98 51.67 0.31
(2) 14 72.55 70.11 2.44
15 76.95 75.68 1.27
16 52.25 54.65 ¡2.40
where ‘Y’ is the predicted response; ‘bo’, ‘bi’ and ‘bij’ are 17 51.58 51.67 0.09
constant coefficients; ‘xi’ and ‘xj’ are the coded inde-
pendent variables and ‘e’ is the random error. The pre-
dicted values were determined from the derived by the methanol/oil molar ratio (F value D 72.86)
equations for each of the response. The contour plots and, lastly, the catalyst (F value D 26.22). All of the
and three-dimensional response surfaces were coefficients were considered in the design to mini-
obtained using these values that were used for the mize the error. Table 5 shows the model coefficients
optimization process. The coefficients of the model for and probability values (i.e. coded values).
one response were found out with multiple regression
analysis using Design Expert Software Version 9.0 (Stat-
Ease, Inc., Minneapolis, MN, USA). The analysis of vari-
ance (ANOVA) was used to check the adequacy of each
model. The quality of fit of the second-order equation
was expressed by ‘R2’ and its statistical significance
was determined by F-test [30]. The p value (probability
value) serves as a tool for checking the significance of
each of the coefficients and their interaction, which
may indicate the patterns of the interactions among
the variables. The smaller the p value, the more signifi-
cant the corresponding coefficient [31].
The coded model was used to generate response Response surface analysis
surfaces for the analysis of the variable effects on %
Figure 3a shows the effects of reaction temperature,
FAME yield:
methanol/oil molar ratio and their reciprocal interac-
tions on % FAME yield using a constant catalyst
Y ð%Þ D 51:67 C 6:89 x1 C 4:13 x2 amount of 2 vol. %. As the temperature increases from
40 to 60 C, % FAME yield increases. The responses
C 13:07 x3 ¡ 3:04 x1 2 C 4:13 x2 2 ¡ 3:92 x3 2 increase as the methanol/oil molar ratio increases from
6:1 to 21:1, while beyond a molar ratio value of 21:1,
there was no significant change in % FAME yield.
C 0:44 x1 x2 C 4:10 x1 x3 ¡ 1:79 x2 x3 (3)
Figure 3b shows the effect of different reaction tem-
perature and amount of added catalyst on % FAME
The ANOVA was used to evaluate the adequacy of yield when a constant methanol/oil molar ratio of
the fitted second-order response surface model by the 15.5:1 is maintained. The % FAME yield increases slowly
least squares method and to assess the goodness of fit. as the amount of catalyst increases from 1 to 3 vol. %.
ANOVA is a statistical technique that subdivides the An increase in temperature enhances the % FAME yield
total variation of a set of data into components associ- in a linear trend and it shows that influence of temper-
ated with specific sources of variation. The regression ature on % FAME yield is more significant in compari-
equation obtained from the ANOVA shows (Table 5) son to that of catalyst amount.
that the R2 was 0.98, indicating a good correlation Figure 3c shows the effect of different methanol/oil
between the observed and predicted values. This is an molar ratio and amount of catalyst on % FAME yield at
estimate of the fraction of overall variation in the data a constant reaction temperature of 50 C. The catalyst
accounted for by the model, and thus the model is amount and methanol/oil molar ratio both effectively
capable of predicting the response within 2% variation. influenced % FAME yield. At any amount of catalyst, an
The R2 value provided a measure of how much of increase in the quantity of added methanol/oil molar
the variability in the observed response values could ratio enhanced the % FAME yield in a linear trend, but
be explained by the experimental factors and their methanol/oil molar ratio has more significant effect.
interactions. The adjusted R2 (0.96) was also very The results showed that the two factors temperature
high which indicates that the model is good. The and methanol/oil molar ratio were the most significant
given statistical tests indicated the chosen model to parameters for influencing the % FAME yield. On the
be satisfactory for predicting the castor oil methyl basis of the data shown in the above figures, we deter-
ester yield within the given range of variables, and mined the optimal values of three factors for maximiz-
also showed that the model was valid for the present ing the % FAME yield. The optimal values for these
study. The smaller the p value for a parameter the factors were determined by resolving the regression
more significant the parameter is, and hence it equation using Design Expert 9. The optimal conditions
reflects the relative importance of the term associ- were selected from within the designated range of
ated with that parameter [31]. methanol/oil molar ratio 6:1–25:1, catalyst amount 1–
The normal probability plot of the residuals follows 3 vol. % and reaction temperature 40–60 C. The optimal
a straight line, indicating a normal distribution of the conditions for % FAME yield, as estimated by the model
residuals (Figure 2b). equation, were as follows: x1 D 25:1, x2 D 3 vol. % and
x3 D 60 C (Figure 3d). Theoretically, these conditions
should give a % FAME yield, i.e. Y, of 75.67%. The %
Table 5. Analysis of variance (ANOVA) for response surface FAME yield obtained in the experiment was 76.95%,
quadratic model. which is very close to the theoretically predicted value.
Sum of Mean F
Source squares df square value p value Model validation
Model 2577.36 9 286.37 51.17 < 0.0001
x1 407.76 1 407.76 72.86 < 0.0001 In order to validate the RSM model, two additional
x2 146.74 1 146.74 26.22 0.0014
x3 1469.85 1 1469.85 262.62 < 0.0001 experiments were also conducted at two sets of operat-
x1x2 1.57 1 1.57 0.28 0.6131 ing conditions. The experimentally obtained % FAME
x1x3 134.15 1 134.15 23.97 0.0018
x2x3 25.63 1 25.63 4.58 0.0696
yield and that predicted by the model are given in
x12 64.69 1 64.69 11.56 0.0114 Table 6. From the table it is clear that the RSM model
x22 119.22 1 119.22 21.30 0.0024 predicts the % FAME yield within acceptable accuracy
x32 107.31 1 107.31 19.31 0.0032
Residual 39.18 7 5.60 (within a deviation of § 5%). Therefore, it is concluded
Lack of fit 38.61 2 19.31 171.24 < 0.0001 that the model may be used to predict the % FAME yield
Pure error 0.56 5 0.11
Totals corrected for 2616.53 16 at other values of three operating parameters within a
the mean pre-specified range. The produced FAME was tested for
R2 D 0.98; R2 (adjacent) D 0.96; Standard deviation D 2.37%
different properties and all data are given in Table 7.
BIOFUELS 7
Figure 3. (a) Three-dimensional (3D) response surface plot for the effect of methanol/oil molar ratio and temperature on % FAME
yield; catalyst amount 2 vol.%; (b) 3D response surface plot for the effect of catalyst amount and temperature on % FAME yield at
constant methanol/oil molar ratio of 15.5; (c) 3D response surface plot for the effect of catalyst amount and methanol/oil molar
ratio on % FAME yield at a constant reaction temperature of 50 C; (d) 3D response surface plot representing the effect of catalyst
amount and methanol/oil molar ratio on the optimum % FAME yield at constant at 60 C temperature.
Application of RSM model with a methanol/oil molar ratio of 6:1–25:1 while vary-
ing the amount of catalyst from 1 to 3 vol. % at three
The developed RSM model has been used to study the
constant temperatures of 40, 50 and 60 C, respectively.
effect of variation in operating parameters on % FAME
At 40 C, % FAME yield does not change appreciably
yield. Figure 4a–c shows the variation of % FAME yield
when the methanol/oil molar ratio is increased beyond
12:1. At higher temperatures (50 and 60 C), % FAME
Table 6. Comparison of model prediction with experimental yield increases with the increase in molar ratio.
values obtained additionally at two sets of operating Figure 5a–c shows the variation of % FAME yield
conditions. with the temperatures 40, 50 and 60 C at varying
Operating conditions % FAME yield amounts of methanol/oil molar ratio (6:1–25:1) at three
Catalyst Methanol/ Value
Temp. amount oil molar Experimental predicted %
constant catalyst amounts between 1 and 3 vol. %,
S. no. ( C) (vol. %) ratio value by model Deviation respectively. The % FAME yield increases following a
1 45 3 12:1 45.15 44.99 ¡0.3543 similar trend for all three catalyst amounts. Further, it
2 55 3 18:1 67.56 70.93 C4.988
8 P. CHAUDHARY ET AL.
Figure 4. (a) % FAME yield with varying molar ratio and catalyst amount at constant temperature of 40 C; (b) % FAME yield with
varying molar ratio and catalyst amount at constant temperature of 50 C; (c) % FAME yield with varying molar ratio and catalyst
amount at constant temperature of 60 C.
Table 7. Properties of FAME from castor oil. is noted that up to 40 C the methanol/oil molar ratio
Standard value/ does not show an appreciable effect on % FAME yield
Determined in range Canoira for all three catalyst amounts, but with further increase
S. no. Property present study et al. [33]
1. Kinematic viscosity 18.55 3.5–5.00
in temperature, % FAME yield increases with the
(mm2/s at 40 C) increase in molar ratio.
2. Density (kg/m3) 920.5 860–900 Figure 6a–d shows the variation of % FAME yield
3. Acid Value (mg KOH/ 0.25 <0.50
g) with catalyst amount 1–3 vol. % at different tempera-
4. Flash point ( C) 170 >120 tures of 40, 50 and 60 C with four constant methanol/
5. Water content (%) 0.006 <0.0005 or <500
mg/kg oil molar ratios between 6:1 and 25:1, respectively. In
6. Calorific value (MJ/ 39.5 37.27–38.22 all four parts of Figure 6, an increase in % FAME yield
kg)
7. Oxidation stability at 8 >6.0 follows the same trend with the increase in catalyst
110 C (hr) amount. It is interesting to note that at a catalyst
8. Sulfur content (mg/ 0.03 <10
kg)
amount of 2 vol. %, the % FAME yield curve passes
BIOFUELS 9
Figure 5. (a) % FAME yield with varying molar ratio and temperature at constant catalyst amount (1 vol. %); (b) % FAME yield with
varying molar ratio and temperature at constant catalyst amount (2 vol. %); (c) % FAME yield with varying molar ratio and tempera-
ture at constant catalyst amount (3 vol. %).
through minima for all of the cases. This phenomenon be within § 5% deviation. Using the RSM model, the
is more predominant at higher temperatures (50 and effect of variation in operating parameters on % FAME
60 C). yield was also studied, as shown in Figures 4a–c, 5a–c
and 6a–d.
Conclusion
Abbreviations
In this study, CCD was used to find out the optimized
operating parameters (reaction temperature, catalyst ANN artificial neural network
amount and methanol/oil molar ratio) for the transes- FAAE fatty acid alkyl esters
terification reaction of castor oil in the presence of FAME fatty acid methyl esters
H2SO4 catalyst, and the optimum value of % FAME CCD central composite design
yield. The optimum values of these three independent GC gas chromatograph
parameters obtained by using CCD were 25:1 methano- RSM response surface methodology
l/oil molar ratio, 3 vol. % catalyst amount and 60 C
reaction temperature. At these optimum conditions,
76.95% FAME yield was observed. The developed RSM
model was also validated with an additional set of
Disclosure statement
experimental data, and its predictions were found to No potential conflict of interest was reported by the authors.
10 P. CHAUDHARY ET AL.
Figure 6. (a) % FAME yield with varying molar ratio and catalyst amount at constant molar ratio 6:1; (b) % FAME yield with varying
molar ratio and catalyst amount constant molar ratio 12:1; (c) % FAME yield with varying molar ratio and catalyst amount at con-
stant molar ratio 18:1; (d) % FAME yield with varying molar ratio and catalyst amount at constant molar ratio 25:1.
removal in feedstock. Appl Biochem Biotechnol. [22] Liu K-S. Preparation of fatty acid methyl esters for gas-
2007;136–140. chromatographic analysis of lipids in biological materi-
[13] Yuan X, Liu J, Zeng G, et al. Optimization of conversion als. JAOCS. 1994;71:1179–1187.
of waste rapeseed oil with high FFA to biodiesel using [23] Verma P, Sharma MP. Comparative analysis of effect of
response surface methodology. Renew Energ. methanol and ethanol on karanja biodiesel production
2008;33:1678–1684. and its optimization. Fuel. 2016;180:164–174.
[14] Rashid U, Anwar F, Ashraf M, et al. Application of [24] Saydut A, Erdogan S, Kafadar AB, et al. Process optimization
response surface methodology for optimizing tranester- for production of biodiesel from hazelnut oil, sunflower oil
ification of Moringao leifera oil: Biodiesel production. and their hybrid feedstock. Fuel. 2016;183:512–517.
Energ Convers Manage. 2011;52:3034–3042. [25] Chaudhary. P, Kumar B, Kumar S, et al. Castor oil transes-
[15] Tiwari AK, Kumar A, Raheman H. Biodiesel production terification: Experimental and modeling studies. Chem
from jatropha oil (Jatropha curcas) with high free fatty Prod Process Model. 2015;10 (2):71–80.
acids: an optimized process. Biomass Bioenerg. [26] Lotero. E, Yijun L, Lopez DE, et al. Synthesis of biodiesel via
2007;31:569–575. acid catalysis. Ind. Eng. Chem. Res. 2005;44 (14):5353–5363.
[16] Freedman B, Butterfiels RO, Pryde EH. Transesterification [27] Kumar S, Singh RK. Optimization of process parameters
kinetic of soyabean oil. J Am Oil Chem Soc. by response surface methodology (RSM) for catalytic
1986;63:1375–1380. pyrolysis of waste high-density polyethylene to liquid
[17] Agra IB, Warnijati S, Wiratni W. Two steps ethanolysis fuel, J Environ Chem Eng. 2014;2:115–122.
of castor oil using sulfuric acid as catalyst to produce [28] Gahdge SV, Raheman H. Process optimization from bio-
motor oil. World Renew Energ Cong. 1996;9:1025– diesel production from mahua (Madhuca indica) oil
1028. using response surface methodology, Bioresour Tech-
[18] Zieba A, Matachowski L, Lalik E, et al. Methanolysis of nol. 2006;97,379–384.
castor oil catalysed by solid potassium and cesium salts [29] Kilic M, Uzun BB, Putun E, et al. Optimization of biodiesel
of 12-tungstophosphoric acid. Catal Lett. 2009;127:183– production from castor oil using factorial design. Fuel
194. Process Technol. 2013;111,105–110.
[19] Cavalcante KSB, Penha MNC, Mendonca KKM, et al. Opti- [30] Montgomery DC. Design and analysis of experiments.
mization of transesterification of castor oil with ethanol 5th ed. New York: John Wiley and Sons; 2001.
using a central composite rotatable design (CCRD). Fuel. [31] Khuri AL, Cornell JA. Response surfaces design and anal-
2010;89:1172–1176. ysis. New York: Marcel Dekker; 1987.
[20] Silva N, Batistella CB, Filho RM, et al. Biodiesel produc- [32] Dias JM, Araujo JM, Costa JF, et al. Biodiesel production
tion from castor oil: optimization of alkaline ethanolysis. from raw castor oil. Energ. 2013;53:58–66.
Energ Fuel. 2009;23:5636–5642. [33] Canoira L, Galean JG, Alcantara R, et al. 2010. Fatty acid
[21] Naik M, Meher LC, Naik SN, et al. Production of biodiesel methyl esters (FAMEs) from castor oil: Production pro-
from high free fatty acid Karanja (Pongamia pinnata) oil. cess assessment and synergistic effects in its properties.
Biomass Bioenerg. 2008;32:354–357. Renew Energ. 2010;35:208–217.