Discrimination of oils and fuels using a portable NIR spectrometer

[Display omitted] •Portable NIR spectrometer was used to discriminate crude oils from used motor oils.•PLS-DA model discriminated crude oils with precision over 94%.•Naphtha was quantified in gasoline with LOD of 4.4 wt%.•Diesel was quantified in kerosene with LOD of 9.3 wt%. Improper mixtures of: motor oil with crude oil; and derivatives mixed with other derivatives of lesser commercial value were identified in Brazil by companies in the energy sector. This study shows the great response that a portable NIR spectrometer had to discriminate crude oils and derivatives and to quantify them in blends (crude oils with used motor oil; and naphtha, gasoline, diesel, and kerosene). NIR spectra set were acquired in triplicate using a microNIR™ portable spectrometer, where it was possible to discriminate crude oil from used motor oil with 100% sensitivity, specificity, and precision. Regression models can quantify the oil content of a ternary mixture containing two crude oils (light and heavy oil) and a used motor oil with root mean square error of prediction (RMSEP) of 6.2 and 4.8 wt%, and R2p = 0.9871 and 0.9870 for support vector regression (SVR) and partial least squares (PLS), respectively. About the NIR spectra of naphtha, gasoline, diesel, and kerosene, partial least squares discriminant analysis (PLS-DA) allows the identification of any of these products with sensitivity, specificity, and precision of 100%. For the blends of gasoline and naphtha, the limit of detection (LOD), limit of quantification (LOQ), and RMSEP were 1.3, 4.4, and 1.4 wt%, respectively. Likewise, for diesel and kerosene blends, the PLS model allows the identification of the diesel with LOD, LOQ, and RMSEP of 2.8 wt%, 9.3 wt%, and 11.4 wt%, respectively.