Prediction of glycerol removal from biodiesel using ammonium and phosphunium based deep eutectic solvents using artificial intelligence techniques

Abstract

Biodiesel total glycerol content is an important characteristic which must pass the EN 14214 and ASTM D6751 international biodiesel quality standards. In this study, the experimental data of glycerol removal by means of deep eutectic solvents (DESs) was used to design a new modeling approach based on Artificial Neural Networks (ANNs) in order to predict glycerol removal. The DESs were synthesized with choline chloride and methyl triphenyl phosphunium bromide as salts and different hydrogen bond donors. DESs composition and the mole fractions of DESs to biodiesel were used as inputs to the model. A feed-forward neural network with 4 hidden neurons was applied and training was done based on the Levenberg–Marquardt optimization method. The ANN prediction was in good agreement with the measured data with an absolute average deviation of 6.46%. The predicted results indicated that the DESs synthesized with glycerol as hydrogen bond donor has lower removal efficiencies. Furthermore, the phosphunium-based DESs were much efficient in attracting total glycerol in comparison with ammonium-based DESs.