Characterization of crude oils with a portable NIR spectrometer

Abstract

The characterization of oils by the physicochemical properties analyses is the basis of the industrial chain. Multivariate data analysis seeks to optimize this process through a characterization database. Thus, it is possible to build different models that estimate all these properties with only a data set from a spectroscopic technique. Near-infrared (NIR) spectroscopy is widely used for this purpose. However, there are no studies in the literature using a portable instrument – a technological advance interesting, for example, in field situations. In this study, we evaluate the performance of a portable NIR spectrometer and compare it with a benchtop NIR instrument. We estimate API gravity (°API), total sulfur content (TSC), basic nitrogen content (BNC), saturates hydrocarbons (Sat), aromatics (Aro), resins (Res), and asphaltenes (Asp) from a set of 182 Brazilian crude oil samples. We constructed partial least squares (PLS) and support vector regression (SVR) models. The PLS Asp model obtained the best response with the portable instrument in the wavenumber range from 11,000 cm−1 to 6,000 cm−1. However, the benchtop instrument was superior for the PLS models of the other properties because it operates in the range from 10,000 cm−1 to 4,000 cm−1. The range from 6,000 cm−1 to 4,000 cm−1 is important for this algorithm because it presents the first overtones of S-H, O–H, C–H, and N–H bonds. The accuracy to predict these properties with the portable instrument can be improved with the application of the SVR algorithm. The results obtained with SVR were similar or superior (Sat, Aro, and Asp) to those obtained with the data from the benchtop instrument. For Sat, Aro, and Res, the SVR models still obtained a coefficient of determination (R2) below 0.7. So, other data treatment must be investigated to improve the accuracy of these properties. Even so, the portable instrument has applicability to petroleum analysis. It showed satisfactory results to estimate the properties °API, TSC, BNC, and Asp.