The economic importance of fuels demand quick response tools to control their qualities, as they are subject to adulteration and/or contamination. In this context, the Partial Least Square Discriminant Analysis (PLS-DA) can be used to classify samples inside and outside a given interest, so its use is proposed for the construction of a multivariate model that discriminates samples of biodiesel blends in diesel from according to their concentrations. The model was built from 68 samples where: 34 simulate crambe methyl biodiesel blends in diesel with a content of 10.0 ± 0.5 (% v/v), our class of interest, and 34 simulate biodiesel-diesel blends in a range ranging from 1.0 to 9.0 (% v/v) and from 11.0 to 30.0 (% v/v), our non-interest class. The classification model shown in Figure 1 presented VP=1, FP=0, VN=1, FN=0 and CCM=1, managing to correctly differentiate the samples of interest and not.
Figure 1 – Samples vs. Classes provided by PLS-DA