Experimental study on the effects of feedstock on the properties of biodiesel using multiple linear regressions

Biodiesel is a very promising alternative fuel that has its place in the future energy mix. The dependence of fuel properties on fatty acids profile will influence the choice of feedstock or appropriate treatment that it should undergo in order to respect biodiesel standards. The objective of this study is to find models that predict biodiesel's viscosity, density, flash point, higher heating value, and oxidative stability based on saponification value, Iodine value and the polyunsaturated fatty acids content of feedstock. Biodiesel samples were produced from seventeen different blends of oils. Multiple linear regressions were used to obtain models. High accuracy prediction was obtained for density and higher heating value with prediction errors <5%, a very good accuracy was obtained for viscosity with error <10% and flash point and oxidative stability were predicted with a fair accuracies (error < 15%) which indicates a good correlation level with IV, SV and Polyinsaturations but it also reveals that other parameters could also interfere and should be taken in consideration to reach acceptable accuracy.