Modeling and Optimization of Transesterification of Beniseed Oil to Beniseed Methylester: A Case of Artificial Neural Network versus Response Surface Methodology

Abstract

In this research work, statistical approach (ANN and RSM) were used to optimize the transesterification of beniseed oil to beniseed methyl ester (BME). Analyses of an heterogeneous catalyst (Mangifera indica powdered) obtained from unripe Mangifera indica peels showed that the powder consist macro elements such as K (59.85%), Si (30.53%), Cl (4.58%), Al (3.05%) and Ca (1.05%) and micro elements such as P (0.196%), S (0.593%), Mn (0.043%), Fe (0.037%), Zn (0.008%), Rb (0.042%) and Sr (0.032%). ANN predicted optimal condition for Beniseed methyl ester produced was X1= 60.0 min, X2 = 1.0 wt.%, X3= 57 0C and X4 = 6.0. The predicted BME (94.40% (w/w)) under this condition was validated to be of 93.80 % (w/w). Meanwhile, RSM predicted 94.20% (w/w) at the following condition X1= 62.0 min, X2 = 0.9 wt. %, X3= 60 0C and X4 = 6.5 was validated as 92.80 % (w/w). The results obtained showed the superiority of ANN over RSM owing to its higher values of predicted value, RMSE, AAD, R2 and R2Adj. The fatty acid profile and the physicochemical properties of the BME indicated that, BME can serve as alternative fuel for conventional diesel.