At the present study, the methalonysis process of sunflower oil was investigated in order to obtain methyl esters (biodiesel) by means of homogeneous alkaline catalyzed transesteriï¬cation.
For the purpose of achieving the maximum process yield of biodiesel comprising the most important fuel properties which satisfies the American standards for testing of materials (ASTM) legislation, central composite design (CCD) was employed (1) to optimize the effect of most influential variables (time of reaction, alcohol and catalyst amounts) on yield of transesterification and (2) to design the experiment respectively. As a result, the optimum sunflower oil methalonysis conditions for obtaining the highest predicted response (100 wt%) elucidated to be : the reaction time of 60 min, excess stoichiometric amount of alcohol to oil of 25% w/w and the catalyst content of 0.5 % w/w .In addition a number of fuel specifications of the resultant biodiesel was tested according to the accepted methods were found to meet nearly all recommended ASTM D 6751 requirements and indicated to be a possible alternative for petro diesel.
1. Introduction
At the recent decades, the high global demand of energy along with the dramatic decrease in petroleum sources attracted the world´s attention to substitute any alternative sources for the conventional petroleum fossil fuels.
Biodiesels are fatty acid alkyl monoesters produced by the reaction of low molecular alcohol (e.g. ethanol, methanol) with triacylglycerols. Since the important fuel properties of the most biodiesels are close to those of conventional diesel fuels, biodiesels can be considered as appropriate replacement for the diesel fuels.
* First author, E-mail address: [email protected]
**Corresponding author. Tel.: +60 3 89466447; fax: +60 3 89464440. E-mail address: [email protected] (T.C. Ling)
Furthermore biodiesels are the sources of energy which is renewable, biodegradable, non-toxic and environmental friendly. Although there are few chemical methods for the conversion of oils and fats into biodiesel (e.g. transesterification, pyrolysis, blending and micro emulsions) the most recommended one is mentioned to be the transesterification(Ghadge & Raheman, 2006; Ma & Hanna, 1999).
Transesterification (alcoholysis) is a method in which a triglyceride molecules (comprising 98% of vegetable oils components) split up during the reaction of a monohydric alcohol with the glycerol part of triglyceride. Therefore the glycerol part would be replaced with alkyl group of monohydric alcohol which leads to mono alkyl esters (biodiesels)formation (Fig.1)(Sanli & Canakci, 2008).The main parameters recognized to have influences on transesterification reaction are the catalyst, alcohol amount, reaction time, temperature, free fatty acids(FFA)and the water content of the reactants(Ghadge & Raheman, 2006).
Fig.1. The transesterification of vegetable oils diagram.
Transesterification can be accelerated by either acidic or basic catalysts however the base-catalyzed reaction is faster than that of acid-catalyzed. Regarding to this reason together with that the acidic catalysts are more corrosive than basic ones, industrial process is usually conducted by using basic catalysts. Strong bases such as potassium hydroxide (KOH), sodium hydroxide (NaOH) and sodium methoxide (CH3ONa) are preferred as strong catalysts. As the type of the catalyst shown to have effect on the rate of phase separation (alkyl ester and glycerol)after reaction completion , sodium methoxide exhibited to be the most active basic catalyst which judged to result in a good phase separation(Hameed, Lai, & Chin, 2009; Korus, Hoffman, Bam, Peterson, & Drown).
Theoretically, in stoichiometric transesterification reaction 1 mol of triglyceride requires 3 mol of alcohol to form 3 moles of alkyl ester (Fig. 1). However, in actual reaction an excess of the alcohol is needed to increase the yields of the alkyl esters production. Furthermore the alcohol type demonstrated to have great effect not only on reaction kinetics but also the fuel characteristics of the resultant product. Based on the result obtained, the yield of biodiesel produced by using methanol is higher than ethanolysis(Hossain, Boyce, Salleh, & Chandran, 2010; Nye, Wllliamson, Deshpande, Schrader, & Snively, 1983).This phenomenon was attributed to formation of stable emulsions between glycerol and alkyl monoester rich layers during ethanolysis(Hossain, et al., 2010; Zhou, Konar, & Boocock, 2003).So the most preferred alcohol was found to be methanol due to its low price as well as chemical properties which can react with triglycerides feasibly and dissolve the catalyst faster than other alcohols(Sanli & Canakci, 2008).Also the alcohol used needs to be necessarily anhydrous as the water is negatively affecting the conversion of oil into esters and in case of water existence in reaction mixture, the transesterification (alcoholysis) would be replaced by hydrolysis and results in formation of free fatty acids instead the alkyl esters afterward(Klok, Robbert, Verveer, & Hendrik, 1993).
Also the presence of high amount of FFAs in feed stock vegetable oil can adversely influence the yield of transesterification by reaction with basic catalysts and soap formation. Due to the formation of soap and water which results in emulsification of the mixture the phase separation stage will be more difficult and costly afterwards .Hence, in case of high levels of FFAs (more than 0.5%) in feed stock an acid catalyzed reaction is recommended (Nye, et al., 1983; Rashid, Anwar, Moser, & Ashraf, 2008; Sanli & Canakci, 2008) .
In many researches, the ambient temperature showed to be enough for transesterification reaction. Although no specific effect on maximum conversion of oil to ester has been detected ,the higher temperatures shown to decrease the maximum conversion time(Korus, et al.).In this study, temperature hasn't been considered to be an independent variable affecting the transesterification reaction. So the value of reaction temperature was supposed to be fixed at 75áµ’ C in order to investigate the effect of the main influential factors.
Sunflower oil is extensively used for frying and cooking purposes it is also going to be attentended as a feedstock for biodiesel production. Besides there is no adequate data on optimization of sunflower oil methalonysis condition using homogeneous alkaline catalyst (NaOCH3).
Therefore, the main objective of this work was to find and focus on the reaction parameters (variables) with significant effect on FFAs conversion to their methyl esters (response) through a simple transesterification reaction. For this reason central composite design (CCD) and response surface methodology (RSM) were employed to predict responses for the sets of experimental design based on CCD matrix in which the parameters varied within their defined ranges to obtain an optimum condition for biodiesel production. In addition a number of important fuel properties of the resultant sunflower oil biodiesels from an optimum condition were assessed and compare with ASTM standards.
2. Materials & Methods:
2.1. Materials
Sunflower oil was purchased from Malaysian markets. Methanol (reagent grade) and n-heptane as a solvent for gas chromatographic (GC) analysis ware purchased from Fisher Scientific (Malaysia). Pure methyl heptadecanoate (≥99.5%) used as a reference internal standard and sodium methoxide as an alkaline catalyst were purchased from Sigma-Aldrich (Malaysia).
2.2. Production of methyl esters
This experiment was carried out using a laboratory scale reactor IKA (LR 2000V, Germany). The system consisted of a double jacket glass vessel equipped with a mechanical stirrer, water condenser, temperature regulator, sampling outlet and an adjustable water bath providing the desired temperatures.
2.2.1. Transesterification
In brief, a fixed amount of sunflower oil was preheated nearly to the set temperature of 75áµ’C.The catalyst completely dissolved in methanol according to the amounts from experimental design and the fresh solution added to the reactor .A vigorous agitation (300 rpm) was begun to proceed the methanolysis reaction in a closed vessel and under a reflux condition for the suppression of alcohol loss.
2.2.2. Phase separation
After the reaction completion, the mixture was transferred into a seperatory funnel and allowed to settle for 1hr during which a complete separation of the product into two distinct layers was achieved. The glycerol as a by-product formed the bottom layer due to its higher density than that of fatty acid methyl esters (FAMEs) and was decanted by the means of the seperatory funnel.
2.2.3. Neutralization
The methyl ester upper layer was then washed four times using a warm mild acid solution (5% citric acid dissolved in distilled water 50-60áµ’C) followed by a warm pure distilled water where the foresaid washing process was repeated for another cycle. In fact utilizing the mild wash warm prevents the formation of immiscible ester/water emulsions as well as the saturated fatty acids sedimentation during the washing procedures. The main purpose of the washing is to decompose the soap formed during the transesterification process while the unreacted tri- di- and monoglycerids and the residual alcohol were also removed(Rashid, et al., 2008; Sanli & Canakci, 2008).Moreover the addition of acid will result in the neutralization and removal of alkaline catalyst. Van Gerpen et al. indicated that any further washing process beyond four times didn't show any additional benefits(Vangerpen, Hammond, Yu, & Monyem, 1997).
2.2.4. Distillation
Subsequently, the methyl ester fraction was distilled off using the rotary vacuum evaporator for 1hr at 80áµ’C under a moderate vacuum for complete elimination and recovery of the residual water and methanol(Klok, et al., 1993; Rashid, et al., 2008).The resulted Fatty acid methyl ester product was then stored under the nitrogen at -18áµ’C before the gas chromatography analysis.
2.3. Analytical methods
2.3.1. Internal standard solution and sample preparation
A solution of methyl heptadecanoate (internal standard) in heptane (10g/mL) was prepared as a calibration for quantification. Approximately 250 mg of each sample (FAMEs) obtained from transesterification were accurately weighed and diluted with 5mL of internal standard solution(Wang & McCurry, Determining the ester and linolenic acid methyl ester content to comply with EN14103).
2.3.2. Gas chromatography analysis (GC)
Sample analysis was performed by gas chromatography using Agilent GC7890 equipped with FID (flame ionization detector) and auto-sampler injector controlled by a PC with software.The injection was done at split mode (100:1) through utilizing a highly polar BPX70 capillary column and 30 m in length. The column temperature program raised from 140 °C to 180 °C at the rate of 10 °C/min then to 220°C at the rate of 2°C/min and holding for 1 min. The inlet and FID temperatures were 230°C and 250°C respectively and the carrier gas was the high purity nitrogen with the flow rate of 0.7 ml/min. The methyl esters yield (Y) in each experiment was calculated using the following formula as a mass fraction in percent Eq.(1)(Ruppel & Huybrighs):
(1)
∑ A = Total peak area
AIS = Internal standard (methyl heptadecanoate) peak area
CIS = Concentration of the internal standard solution (mg/mL)
VIS = Volume of the internal standard solution (mL)
M = Mass of the sample (mg)
2.4. Design of experiment
Response surface methodology (RSM) was applied to design the matrix of experiments(Table.1 ) to investigate the effect of three main independent variables namely the reaction time (60 -180 min, x1), excess stoichiometric amount of methanol to oil (25-125 wt%, x2) and the catalyst concentration (0.1-0.9 wt%, x3)on transesterification yield (wt%,Y )of sunflower oil As a response.
The type of catalyst, agitation rate and the reaction temperature were kept fix through the experiments so twenty transesterification experiments were designed based on central composite design (CCD) which is suitable for the optimization of influential factors with a minimum number of experiments. The designed matrix included three independent variables with five levels for each factors in which the method repeatability was calculated through repeating the center point for six times and the sequences of experiments was randomized for the sake of minimizing the effects of any uncontrolled factors on response (Montgomery, 2001).
2.5. Statistical analysis and model fitting
As for the experimental design, analysis of data was also accomplished by response surface methodology using the Minitab v.14 statistical package (Minitab Inc., PA, USA).The analysis procedure was 1) to specify regression coefficients as well as significant model terms recognition 2) fitting the regression model to determine the optimum levels of the factors leading to an optimal region for the response.
Since the response variable was transesterification yield (Y) an empirical regression model was employed to better understanding the correlations between the reaction factors with response using a second-degree polynomial equation that can be written generally as the followings Eq.(2)(Montgomery, 2001):
Where Y is the predicted response, β0 the offset term, βi the linear coefficients, βii the quadratic and βij are the interaction coefficients and xi and xj are the independent variables.
Since there were three factors involved in this study ( the mathematical relationship between factors and response becomes Eq.(3):
(3)
The response surface models adequacy was tested by analyzing the models determination coefficients (R2)(Weng, Liu, & Lin, 2001).It was suggested that for a good fit of model , R2 shouldn't be less than 0.80 (Joglekar & May, 1987).Therefore ,The factors possess the smaller p-values (p0.05) and the larger magnitude of t-values have the higher corresponding coefficients and the higher significant effect on response variable.
Hence, the terms with non-significant statistical effect (p were eliminated from the initial regression model and the experimental data were refitted with only significant factors (pto achieve a final reduced model. It should be noted that where the linear terms of some independent variables are non-significant (pthey still may be kept in the model, since the interaction and quadratic terms containing the foresaid terms (p0.05)(H. Mirhosseini, Tan, Hamid, & Yusof, 2008) .
2.6. Optimization and validation
The optimization process was performed according to response surface methodology procedures (Minitab software) for the prediction of statistical optimum levels of three involved independent variables (x1, x2 and x3) which subsequently resulted in the desired response goals.
The numerical optimization was accomplished by response optimizer to obtain the individual and overall desirability along with searching for the exact values of independent variables´ combination point that jointly leads to optimize a set of responses through satisfying the requirements in each response. In this study since there is only one response investigated which is the sunflower oil trans-esterification yield the individual desirability (d) become equal to that of overall (D)(Minitab Inc., 2000).
The graphical optimization was performed utilizing three dimensional (3D) surface plots for a better visualization of independent variables (Alcohol, Time and Catalyst) effect on response variable in reduced fitted model. The 3D plots were created by keeping one variable as constant at its center point and varying two other factors within their experimental ranges. As shown in 3D plots the interaction terms of alcohol/time (Fig.2) and alcohol /catalyst (Fig.3)haven't exhibited any significant effect on transesterification yield. Besides, for checking the adequacy of regression model the actual experimental data were compared to predicted values calculated by the software(Fig. 4)(Hamed Mirhosseini, Tan, Taherian, & Boo, 2009).
3. Results and discussion
The experimental data together with the predicted values are shown in Table.2 and the regression coefficients, F-ratio and p-value of the reduced model for linear, quadratic and interaction effect of the terms which conducted any significant influences (p≤ 0.05) on response variable were specified in Table.3. Aforesaid table also included the coefficient of determination (R2) and adjusted R2 with the values higher than 0.80 indicating the adequacy of the model.
As shown in table 3 , the single effect of the reaction time (x1) and the catalyst concentration (x3) along with the quadratic effect of excess stoichiometric amount of methanol to oil (x2* x2 ) played significant roles on the variability of sunflower oil transesterification yield (Y ) .
Consequently, the terms except for the mentioned terms with high significancy were dropped from the initial model achieving a fitted final model. However the main term of excess stoichiometric amount of methanol to oil (x2) was still remained in the final model in favor of the quadratic term significant effect.
3.1. Effect of reaction time(x1)
The recorded results in Table.3 showed that the transesterification (methylation) yield was profoundly influenced by the reaction time (p<0.05).Subsequently the linear term of the reaction time which displayed the lowest p-value besides the highest F-ratio concluded to be the most significant term among others.
However, according to estimated coefficients reported in Table.3, the reaction time negatively affected the transesterification yield meaning that when the reaction time decreased the greater amount of fatty acid methyl esters (FAMEs) would be produced. This matter can be attributed to the ester hydrolysis after the formation (Sanli & Canakci, 2008).As a result the response surface optimizer suggested the reaction duration of 50 minutes in which the predicted response yield would be maximized.
3.2. Effect of alcohol(x2)
As mentioned earlier, excess stoichiometric amount of monohydric lower-alkyl alcohol with respect to the fatty acid residues in one or more fatty-acid glycerol esters is used to increase the yield of transesterification with regards to the volatile nature of the alcohol (Klok, et al., 1993).
As observed in Table.3, there is no significant correlation between the linear term of alcohol(x2) and methylation yield (p.However, the quadratic term of alcohol (x2* x2) significantly affected the methylation yield as the response variable (p0.05). Hence, the response surface optimizer predicted that to achieve the highest yield in methylation the amount of alcohol in oil (%wt) should be 25% more than that of stoichiometric amount. As shown in Fig.2 excess amount of alcohol higher than 25% up to75% exhibits an intensive drop in methylation yield however at the range of 75-125% increasing the alcohol content lead to methylation yield enhancement. Since this result was in accordance with previous studies (Antolin, et al., 2002)it´s necessary to keep the methanol to oil ratio efficiently low to 1) optimize the phase separation of fatty acid methyl esters and glycerol 2) decrease the environmental contamination and economical costs.
3.3. Effect of catalyst (x3)
Based on the results in Table.3 the linear term of alkaline catalyst(sodium methoxide) significantly affected the methylation yield (p0.05).As observed in aforesaid table the corresponding estimated coefficient demonstrated a positive correlation between the linear term of catalyst concentration and the methylation yield concluding that the rate of transesterification could be accelerated by using the catalysts. Accordingly, through the optimization results the highest yield of methylation was predicted when the catalyst concentration was 0.5 %wt of the oil. As a matter of fact the optimum predicted amount of catalyst was efficient and increasing the amount not only isn't necessary to enhance the methylation yield but also adds extra costs to transesterification process.
3.4. Validation of the final reduced model
In order to prove the adequacy of response surface optimization model the experimental values were compared to predicted values computed from the software. As shown in Table.2 no significant differences were observed between the experimental and predicted values (p0.05). Also, the closeness of actual and predicted values showed in Fig.4 confirming that the regression equation is adequate (H. Mirhosseini, et al., 2008) . Accordingly the desirable fatty acid methyl ester mixture which was predicted to contain the highest percentage of methylated fatty acids (FAMEs) was empirically produced under the optimum condition and subsequently the GC analysis confirmed that there is no significant difference between the FAMEs produced under the optimized condition and the experimental value.
4. Conclusion
In this case, response surface methodology was utilized to investigate the effect of process conditions on response variable. Generally CCD was exhibited to be a beneficial design that efficiently specified of the influences of three independent variables on response variable. Subsequently the regression equation was efficiently significant with p>0.05 and R2> 0.8 showing indicating that 89% of the total variation in transesterification yield was attributed to the experimental independent variables. Therefore, the best combination of factors for obtaining the highest predicted response (100 wt%) are the reaction duration of 60 min, extra alcohol amount of 25% w/w and catalyst content of 0.5 % w/w .Since the catalyst is needed in minute amounts, the primary cost would be only relevant to the alcohol. Consequently as the use of alcohol can be reduced without significantly reducing the transesterification yield, the biodiesel production cost would be lowered and the process would be more economical.
Hence by optimization of the process the yield increased to the highest value. Furthermore, the results of quality assessment tests performed on the final product (Table.4) demonstrated that fuel specifications were in agreement with ASTM D 6751 requirements meaning that the biodiesel obtained by this process conditions is of good quality and suitable for utilization in automotive engines(ASTM).
Fig.2.Transesterification yield vs. alcohol and time.
Fig.3.Transesterification yield vs. alcohol and catalyst
Fig.4. Fitted line plot displayed the correlation between the predicted (Yi) and experimental (Y0) methyl ester (biodiesel) contents.
Table.1.Central composite design (CCD) matrix
Treatment
run
Blocks
Time of reaction
(x1,min)
Excess stoichiometric
amount of alcohol to oil (x2,%w/w)
Catalyst
amount
(x3,w/w%)
1C
2
120
75
0.5
2
2
150
50
0.3
3
2
90
100
0.3
4C
2
120
75
0.5
5
2
90
50
0.7
6
2
150
100
0.7
7C
1
120
75
0.5
8
1
150
50
0.7
9
1
150
100
0.3
10
1
90
50
0.3
11C
1
120
75
0.5
12
1
90
100
0.7
13
3
120
25
0.5
14C
3
120
75
0.5
15
3
60
75
0.5
16
3
120
125
0.5
17
3
120
75
0.9
18
3
180
75
0.5
19
3
120
75
0.1
20C
3
120
75
0.5
c Center point.
Table.2. Experimental and predicted values for response variables of reduced model
Treatment run
Blocks
Biodiesel yielda(Y, wt %)
Experimental value
(Y0)
Predicted value (Yi)
1
2
96.46
96.645
2
2
95.4
95.468
3
2
97.32
97.611
4
2
96.2
96.645
5
2
99.2
99.21
6
2
98.46
97.461
7
1
96.95
97.47
8
1
98.14
98.089
9
1
97
96.49
10
1
99.2
98.239
11
1
96.8
97.47
12
1
99.9
100.232
13
3
96.64
96.996
14
3
96.1
95.044
15
3
96.29
96.633
16
3
96.99
97.317
17
3
96.72
96.511
18
3
92.4
93.456
19
3
93.54
93.578
20
3
95.9
95.044
a No significant differences(p˃0.05)between experimental(Y0)and predicted values (Yi).
b Y0-Yi: residual
Table.3. Regression coefficients, R2, adjusted R2, F-ratio and p-value of the final reduced model
Parameter
Model term
Coefficient estimate
F-ratio
p-value
β0
Intercept
96.386
214122
0.000
Linear
β1
x1
-0.794
24.512
0.000
β2
x2
0.0803
0.251
0.625a
β3
x3
0.7332
20.875
0.001
Quadratic
β11
x1* x2
―
―
a
β22
x2* x2
0.528
16.24
0.001
β33
x3 * x3
―
―
a
Interaction
β12
x1* x2
―
―
a
β13
x1* x3
―
―
a
β23
x2* x3
―
―
a
R2
―
0.889
―
―
R2(adj)
―
0.837
―
―
Regression
(F-ratio; p-value)
―
―
239.32
0.000
a Not significant
1, Time of reaction (min); 2, Excess of stoichiometric amount of alcohol to oil (%w/w)
; 3, Catalyst amount (%w/w)
Table.4. Fuel properties of the final optimum fatty acid methyl ester production
Fuel properties
Unit
Value
Standard limits
Standard methods
Specific gravity at 15 áµ’C
―
0.87
0.86-0.90
ASTM D 287
Kinematic viscosity at 40 áµ’C
m2/s
4.110-6
3.510-6- 510-6
ASTM D 445
Flash point
áµ’C
150
≥100
ASTM D 93
Pour point
áµ’C
0
―
ASTM D 97
Ash content
%w/w
0.008
≤0.01
ASTM D 482
Cloud point
áµ’C
2 áµ’C
-1
ASTM D 2500
ASTM; American Standards for Testing of Materials
