LAUSR

Abstract In this study, a novel modeling approach (artificial neural networks (ANN) and ant colony optimization (ACO)) was used to optimize the process variables for alkaline-catalyzed transesterification of CI40CP60 oil mixture (40 wt% of Calophyllum inophyllum oil mixed with 60 wt% of Ceiba pentandra oil) in order to maximize the biodiesel yield. The optimum values of the methanol-to-oil molar ratio, potassium hydroxide catalyst concentration, and reaction time predicted by the ANN-ACO model are 37%, 0.78 wt%, and 153 min, respectively, at a constant reaction temperature and stirring speed of 60 °C and 1000 rpm, respectively. The ANN-ACO model was validated by performing independent experiments to produce the CI40CP60 methyl ester (CICPME) using the optimum transesterification process variables predicted by the ANN-ACO model. There is very good agreement between the average CICPME yield determined from experiments (95.18%) and the maximum CICPME yield predicted by the ANN-ACO model (95.87%) for the same optimum values of process variables, which corresponds to a difference of 0.69%. Even though the ANN-ACO model is only implemented to optimize the transesterification of process variables in this study. It is believed that the model can be used to optimize other biodiesel production processes such as seed oil extraction and acid-catalyzed esterification for various types of biodiesels and biodiesel blends.