Desenvolvimento de modelos quimiométricos baseados em espectroscopia no infravermelho próximo portátil para quantificação de óleo vegetal hidrotratado (HVO) em misturas com biodiesel e diesel de petróleo

Abstract

Hydrotreated vegetable oil (HVO), also known as renewable diesel, has proven to be a viable alternative to the use of petroleum derived fuels. The reference method used to quantify this fuel in petroleum diesel, ASTM D6866-18, based on the measurement of radiocarbon content, is complex and relatively expensive. In addition, it quantifies the total content of bio-based fuels in a blend, not quantifying individual contents of different biofuels. Therefore, the objective of this work was to develop a direct, simple and low-cost analytical method for the quantification of renewable diesel HVO in blends with biodiesel and petroleum diesel, using portable near infrared spectroscopy (NIRS) associated with the multivariate partial least squares (PLS) calibration method. Samples of HVO, petroleum diesel, and biodiesel were supplied by Vibra Energia S. A. (São Paulo, Brazil). Blends were prepared with different proportions of the three components and analyzed using the equipment MicroNIR® 1700. The obtained spectra were preprocessed using standard normal variation (SNV) and mean centering. The samples were divided into calibration and validation sets in two different ways, designed calibration and the algorithm Kennard-Stone, and a model was built with the samples separated from each way. Models with selected variables based on variable importance in projection (VIP) scores, and biodiesel quantification models built with designed calibration were also tested. The models were evaluated by estimating linearity, accuracy, precision, inverse of analytical sensitivity, bias, detection and quantification limits, and robustness. The lowest root mean square errors of calibration (RMSEC) and prediction (RMSEP) for HVO quantification were 0.93% (v/v) and 0.81% (v/v), respectively, obtained by the model built with samples selected by the Kennard Stone algorithm and selected variables. For biodiesel quantification, the lowest values were obtained with the model built with all variables (RMSEP 0.38% (v/v) and RMSEC 0.29% (v/v)). All models were considered accurate, linear, and robust to variations in room temperature and type of agitation during mixture preparation. No systematic errors were detected. The developed method proved to be a viable alternative to reference methods. The results obtained were adequate, based on the estimated figures of merit, and the method is simpler, faster, environmentally friendly (does not consume reagents/solvents) and of lower cost compared to the reference method ASTM D6866-18. In addition, the analyses can be performed at the site of sample use.