Aplicação de espectroscopia no infravermelho médio e calibração multivariada para quantificação de adulterantes em biodieseis de óleo de soja e residual e suas misturas com diesel

Abstract

The National Agency of Petroleum, Natural Gas and Biofuels (ANP) establishes the obligation of adding 5% (v/v) of biodiesel to diesel sold in the commercialized country. During the production residue can be generated from the alcohol used, and this product is undesirable to the quality of biodiesel. During the commercialization an adulteration can occur by the addition of vegetable oils. In both cases it becomes crucial analytical control of this fuel. In this work methodologies were developed and validated based on the use of Spectroscopy in Infrared Middle (MID) combined with multivariate calibration Partial Least Squares (PLS) for quantification of methyl and ethyl alcohols in biodiesels from soybean and residual oil and refined oil and residual in their binary mixtures with diesel. For models of biodiesel / diesel (B5) the results were satisfactory with RMSEP (Root Mean Square Error of Prediction), below 0.27 % (w/w) and MS (mean of relative error) below 1.84 %. Additionally, it was shown an excellent correlation between the actual values and concentration predicted by PLS models, with correlation coefficient (R) above 0.9997 and the absence of systematic errors, based on criteria established ASTM E1655 - 05. As to the inverse of the sensitivity concentration values below 0.09% (w / w) were found. As to the limit of detection and quantification models constructed can only detect than or equal to 0.26% (m / m) and quantified values above 0.79% (w / w). For models of pure biodiesel (B100) the results were also satisfactory with RMSEP below 0.02% (w / w) and below 3.00%. Showed correlation coefficients (R) above 0.9961 and the absence of systematic errors, based on criteria established by ASTM E1655-05. As to the inverse of the sensitivity concentration values below 0.005% (w / w) were found. To the limits of detection and quantification models constructed can only detect than or equal to 0.03% (w/ w) and quantified values above 0.09% (w / w). Therefore, the methodologies developed can be a promising alternative for the quality control of this fuel against possible adulteration.