Characterization of crude petroleum by NIR

Petroleum is a product whose composition (quality) varies due to its natural origin and to transport and storage conditions. In Brazil in particular, crude petroleum comes from a variety of very different oil fields that produce predominantly heavy petroleum. This work proposes a methodology for the estimation of the SimDis (Simulated Distillation) properties of crude petroleum based on NIR (Near Infrared) spectroscopy. SimDis analysis, which takes about three hours to perform, is rapid relative to TBP (True Boiling Point) determination. In the proposed methodology, a NIR transmission probe is used and no desalting or thermal conditioning is applied to the samples before measurement. In order to build the correlations, forty petroleum samples with API grades ranging from 31.1 to 36.4 were employed. The NIR spectral data were correlated with the SimDis curves (weight percent vaporized at a given temperature) and the salt content of the petroleum by neural network techniques. The resulting network reproduced quite accurately the SimDis curves, showing the technical feasibility of the proposed methodology.