Artificial neural network modeling to predict biodiesel production in supercritical methanol and ethanol using spiral reactor

Non-catalytic biodiesel production in supercritical methanol (SCM) and supercritical ethanol (SCE) was conducted using spiral reactor. The experimental data were used to create artificial neural network (ANN) model in order to predict biodiesel yield. The results showed that ANN was the powerful tool to estimate biodiesel yield that was proven by a high value (0.9980 and 0.9987 in SCM and SCE, respectively) of R and a low value (2.72x10(-5), 1.68x10(-3), and 2.30x10(-3) in SCM and 2.24x10(-4), 4.49x10(-4), and 5.03x10(-4) in SCE for training, validation, and testing, respectively) of mean squared error (MSE). For biodiesel production in SCM, the highest yield of biodiesel was determined of 1.01 mol/mol corresponding to the actual biodiesel yield of 1.00 mol/mol achieved at 350 degrees C, 20 MPa within 10 min; whereas, for SCE, the highest yield of biodiesel was observed of 0.97 mol/mol corresponding to the actual biodiesel yield of 0.96 mol/mol achieved at 400 degrees C, 20 MPa within 25 min. (C) 2015 The Authors. Published by Elsevier B.V.