Optimization of Reducing Ffa Content of Waste Cooking Oil by Using Response Surface Methodology

Waste cooking oil has high free fatty acid (FFA). It impact to low yield of biodiesel production. Thus, reducing FFA is one of important process as feedstock of biodiesel. This study aims to investigate the optimum condition of three important process variables which are acid concentration, molar ratio of methanol and oil, and irradiation time with the 45 o C of irradiation temperature for reducing FFA. The synthesis is assisted by ultrasonic irradiation. It conducted by acid esterication with H 2 SO 4 and methanol. Optimization is conducted by Response Surface Methodology (RSM) with central composite design (CCD). The optimum condition of response for reducing FFA less than 1% were found to be 7.22:1 of methanol to oil molar ratio, 0.92% wt of H 2 SO 4 , and 26.04 minutes of irradiation time. It has been observed that ultrasonic system reduces FFA content signicantly compared to conventional method.


Introduction
Biodiesel production is more attractive as fuel due to its feedstock availability and environmental issue. Biodiesel is considered to decrease the dependence to fossil fuel. Waste cooking oil is potential feedstock to be developed since it is non-edible oil and large availability. Thus, it has potential for mass production [1]. However, it has problems to be biodiesel feedstock due to its free fatty acid content. It is reduced by acid esteri cation. Esteri cation is the reaction of two immiscible phases. The less dense phase dissolves the catalyst in the alcohol, while the other phase contains oil. The reaction between the two immiscible phases occurs in the area of the interface between the liquids. Interface area between phases should be increased by vigorous mixing [2]. As we all know, ultrasonic radiation is a means of emulsifying immiscible liquids. This study is conducted by ultrasonic irradiation. It produces bubble cavitation around boundary phase between the alcohol phase and the oil. Emulsi cation is generated during rupture of cavitation bubbles that break boundary phase. Ultrasound impacts one liquid on another liquid [3]. Temperatur will increase locally at boundary phase due to cavitation, thereby transesteri cation reaction enhances signi cantly [4][5] Response surface method (RSM) is a powerfull statistical method that has been carried out in many studies [6]. Multiple regression and correlation analysis are used as tools to evaluate the in uence of two or more independent factors toward dependent variables. In addition, the central composite design (CCD) has been optimized in some chemical processes and biomass technology. The main advantage is to reduce required experimental numbers to provide enough information to obtain acceptable ndings statistically. It has been successfully carried out to optimize the production of biodiesel in oil feedstocks, including Madhuca indica, Jatropha curcas oil, and animal fats [7][8][9][10].
In the recent study, esteri cation based acid is used to reduce high free fatty acid (FFA) of waste cooking oil. Optimization of process variable to be less than 1% using RSM for design of experiments.

Materials
Waste cooking oil was purchased from fried chicken restaurant in Padang, West Sumatera, Indonesia. Methanol and H 2 SO 4 were supplied from Systerm. Its characteristics are reported in Table 1. Approximately 12 g of oil was poured in the round bottom ask. It was heated 45 o C and irradiated by ultrasonic. Subsequently, the certain amounts of methanol and sulfuric acid were poured to the oil.
Temperature was kept constant during esteri cation process. Heating and ultrasonic irradiation were stopped after irradiation has reached the irradiation time. The ask was immersed in cold water immediately for stopping the irradiation. The mixture was stand until separate two layers. The bottom layer was drained whereas the upper layer was washed by hot water for removing the impurities then dried for further analysis (% FFA).

Experimental Design
This study employed RSM with Central Composite Design (CCD). Three process variables are set as independent variables. Those are concentration of sulfuric acid (% H 2 SO 4 ) (C), molar ratio of methanol to oil (M), and irradiation time (t). Dependent variable is free fatty acid (%FFA). A ve-level-three-factors CCD was carried out in this study during 20 experiments (2 k + 2 k + 6). k is the number of independent variables. The level of uncoded and coded (actual) of independent variables are described in Table 2.

Analysis Statistically
Second-order polynomial was applied in analysis as shown in Eq. 1: Where y is the response (percentage of free fatty acid, % FFA); is intercept, is linear constant coe cients, is quadratic constant coe cients, and is interaction constant coe cients. x i and x j are the uncoded independent variables; e is the error. Design Expert 6.0.10 (STAT-EASE Inc) is applied to examine analysis of regression and analysis of variance (ANOVA). Validation of equation is conducted by con rmatory experiments using combinations of independent variables. It is within the experimental area but not part of the original experimental design [7]. The coe cient of determination (R 2 ) is used to evaluate the quality of the model t, and the response surface graph is drawn using the tted quadratic polynomial equation obtained through regression analysis.

Results And Discussion
Model Fitting and ANOVA Table 3 shows the experimental and predicted values of the free fatty acid response percentage obtained at the design point. All variables are displayed in coded and non-coded form.  Table 4.  Table 5. The model F value of 128.75 indicates that the model is valid, and the p value of the model is less than 0.0001, which indicates that the model term is very important in predicting the response value and inferring the applicability of the model. The lack of t is the weighted sum of the squared deviations between the average response of each parameter level and the corresponding t. The p-value for lack of t is 0.1494, indicating that it is not signi cant relative to pure error. The inconspicuous t is not good.
The tted F value is 2.70, which means that when the model is tted to the observed experimental data, the possibility of such a large under tting due to noise is 14.94%. CV of the model is 0.97 that closer to unity. It indicated reliability of tted model is high. The quality of the model t was examined by the coe cient of determination (R 2 ). The R 2 value is between 0 and 1. More closer to 1 indicated reliable model.
The study obtained R 2 0.9914. It shows that 99.14% of the experimental data is compatible with model. The adjusted determination coe cient (adjusted R2) value is 0.9837, which is close to R2. It shows the experimental has strong correlation to predicted values and explains any changes in the response.
Normality plot of data between student residuals and residual are shown in Fig. 2.
It shows that there is a characteristic dispersion of constant variables in the data. The model adequately explains the experimental range studied.
Interaction of Parameters to % FFA Figure 3a shows the effect of %H 2 SO 4 and molar ratio of methanol and oil at 20 minutes of irradiation time and 45 o C of temperature. Percentage of H 2 SO 4 has more signi cant effect to reducing FFA compared to methanol and oil ratio. Interaction between %H 2 SO 4 and molar ratio of methanol and oil has positive effect and signi cant to the reducing FFA content. FFA content decreased with increasing of %H 2 SO 4 and molar ratio of methanol and oil. However, for achieving FFA content less than 1%, employing H 2 SO 4 is less than 1.18% and molar ratio of methanol is more than 7.23:1. Figure 3b represents the effect of irradiation time at 1%H 2 SO 4 and molar ratio of methanol and oil at 45 o C of temperature. Irradiation time has more signi cant effect to reducing FFA compared to molar ratio of methanol and oil. According to plot, FFA content decreased with the increasing molar ratio of methanol and oil. Whereas Fig. 3c represents the effect of %H 2 SO 4 and irradiation time. In interaction between %H 2 SO 4 and irradiation time, irradiation time has more signi cant effect to the reducing the FFA content compared to %H 2 SO 4 .

Process Optimization
Software design expert 6.0.10 is applied to examine process optimization by solving the regression equation (Eq. 1). The model is used to examine the process variable with the smallest FFA content. The optimized result at 45 o C is that molar ratio of methanol and oil is 7.22:1, 0.92 wt% H 2 SO 4 and 26.04 minutes of irradiation time. The model predicts that the lowest FFA content obtainable under these optimal conditions is 0.5%.

Veri cation of Predictive Model
Optimum response value was tested to verify model predicted value. It has been examined to be the optimum response through the RSM optimization method, and is also used to verify the experiment and use the model equation to predict the response value. Table 6 shows the predicted and experimental response values under the best conditions.

Conclusion
In summary, the Response Surface Methodology (RSM) with central composite design (CCD) is successfully conducted to the model to optimize the independent variables for acid esteri cation using ultrasonic irradiation. Ultrasonic irradiation is an effective method to reduce FFA content and save time.
The effect of irradiation time was more signi cant compared to % H 2 SO 4 and molar ratio of methanol and oil whereas %H 2 SO 4 was more signi cant compared to methanol to oil molar ratio in reducing FFA content by acid esteri cation process. RSM generated reliable model in predicting the FFA content precisely. Further, it generated the optimum value for independent parameter. Those were molar ratio of methanol and oil of 7