Performance assessment of artificial neural network on the prediction of Calophyllum inophyllum biodiesel through two-stage transesterification

Bayesian regularized Artificial Neural Network (ANN) coupled with genetic algorithm was used to develop a model to predict the optimized process variable parameters for the transesterification process of the extracted Calophyllum inophyllum bio-oil. In this study, a central composite rotatable design with 27experimental trials by varying the process operating parameters namely, methanol to oil molar ratio, catalyst concentration, and reaction duration are applied to optimize the biodiesel yield. ANN tool predicted the process parameters as 0.94 v/v methanol to oil molar ratio, 0.98 wt% catalyst concentration and 100 min reaction duration to yield a maximum biodiesel of 98.5%. Moreover, the statistical performance indicator of the ANN model showed R, R-2, MSE and MPRD values as 0.97709, 0.98214, 0.13240 and 0.23487, respectively, withhigher precision and accuracy. The optimized process parameters obtained by ANN-GA model was confirmed by conducting further trials based on two-stage transesterification process, and its efficacy was validated with the results of ANN. The physio-chemical properties of the biodiesel were found to be within ASTM D6751 standards.