Prediction of properties of biodiesel-diesel blends using spectrofluorimetry and multivariate calibration

Abstract

Regressions based on fluorescence spectroscopy were developed to provide relatively inexpensive and rapid measurements of the concentration, viscosity, and specific gravity of biodiesel-diesel blends. The methods involved obtaining a mathematical model from spectrofluorimetric data and data from a given property (concentration, dynamic viscosity, or specific gravity) using partial least squares (PLS) regression, which was then applied as a model for predicting properties of interest. The predicted concentrations, dynamic viscosities, and specific gravities of the biodiesel-diesel blends were compared with actual values and agreed reasonably well with the obtained results. The models showed high correlation between real and predicted values. The R-square values near 1 indicated excellent model accuracy for predicting concentrations, specific gravities, and dynamic viscosities of biodiesel-diesel blends. The residual distribution did not follow a trend with respect to the predicted variables, indicating an excellent fit to the data.