QUANTIFICATION OF CRAMBE BIODIESEL CONTENT IN DIESEL BLENDS USING MIR SPECTROSCOPY END VARIABLES SELECTION.

摘要

Mid-Infrared absorption spectroscopy associate with the Partial Least Squares regression is the official method to monitor Brazilian commercial diesel quality. This method, however, uses solvents and a large number of samples for the construction of the calibration curves, which generates waste and increases the time needed for the analysis. In order to develop a non-destructive method, being possible to recover the sample after its quantification, decrease the quantity of samples and make use of a single calibration curve, in this study we used the oilseed crambe, which is not used in human food, for the methyl biodiesel production and PLS analysis for their content determination in mixtures with diesel at a concentration range of 1.00 to 30.00 (% v/v). The global model for crambe methyl biodiesel obatined RMSEC = 0.26 (% v/v), RMSECV = 0.35 (% v/v) and RMSEP = 0.41 (% v/v). Complementary, variable selection method iPLS was applied in the global model in order to reduce the spectral range required to regression construction and to improve the RMSEP, RMSECV and RMSEC values.