Artificial neural network approach for modeling of ultrasound-assisted transesterification process of crude Jatropha oil catalyzed by heteropolyacid based catalyst

Transesterification of crude Jatropha oil to fatty acid methyl esters in an ultrasound-assisted process was conducted in the presence of different heteropolyacid-based catalysts. Tungstophosphoric acid immobilized on activated carbon and gamma alumina as well as cesium salt of the heteropoly acid were prepared and characterized for elucidation of their properties. The experimental data collected from the central composite design were used to establish artificial neural network (ANN) model in order to predict the response in the reaction. The models were also optimized to identify the suitable network topology and training method. The results obtained from ANN models were compared with the results of the regression analysis and good agreement was obtained to suggest the good potential of ANN in the FAME yield prediction.