Near Infrared Spectroscopic Prediction of Physical Properties of Aromatics-Rich Hydrocarbon Mixtures

Abstract

Near infrared spectroscopy multivariate calibration models were used to predict physical properties of aromatics-rich hydrocarbon mixtures obtained from Pyrolysis Gasoline units at the Copene petrochemical plant in Brazil. Relative density and distillation parameters (10%, 50%, 90% and 100% evaporated) were estimated from FT-NIR spectra in the 1670–1800 nm and 2100–2500 nm regions. Principal component analysis was used for exploratory data analysis and cross-validated one-block and two-block partial least-squares models were employed for calibration. Root Mean Square Errors of Prediction, based on six to nine PLS factors, were close to the measured repeatability values and the limits defined by ASTM methods: relative density = 0.0005, 10% evaporated = 0.489, 50% evaporated = 0.350, 90% evaporated = 0.579, 100% evaporated = 3.276. The present work has also demonstrated the ability of chemometric techniques to differentiate between spectra of quite similar aromatic compounds mainly due to the nature and location of substituent groups in the aromatic ring.