Application of near infrared spectroscopy to predict the average droplet size and water content in biodiesel emulsions

NIR spectroscopy was used to predict the average droplet size and water content in water/biodiesel emulsions. The emulsions were prepared from industrial biodiesel obtained from soybean oil (85 wt%) and animal fat (15 wt%) by methylic route. NIR spectra was collected in the transmittance mode with the diffuse reflectance technique. Based on the NIR spectroscopy results, it can be pointed out that this methodology has the sensibility to infer the droplet size and water content in the biodiesel emulsions. Two techniques were used to obtain the multivariate models: the partial least squares (PLS) and artificial neural network (ANN) models. Satisfactory values of mean error for the external validation were obtained, with 9.53% to PLS model for average droplet size and 8.79% for water content, since they are close to the experimental standard deviation. The performance of ANN models demonstrated that this technique allowed the prediction of average droplet size and water content with mean errors of 6.10% and 13.20%, respectively. These errors are close to the analytical error associated to the method used indicating that the NIR spectroscopy is a good alternative to be used for this purpose.