Bromine number prediction for Colombian naphthas using near-infrared spectroscopy and chemometric methods

Abstract

Thirty-eight naphtha samples were used to develop a chemometric method to predict bromine number. All samples were characterized by Fourier transform near infrared spectroscopy (FT-NIR), and their spectra were correlated by similarity using principal component analysis (PCA). The models for bromine number determination (BN) were established by Partial Least Squares regression (PLS) and Multiple Polynomial Regression (MPR). PCA allowed classifying the samples into the light and heavy, determining the most significant spectral variables. These variables are located in the regions between 4000-4800 and 5200-6350cm-1. The results determined by combining FT-NIR spectroscopy and chemometrics were very close to those obtained by standardized methods. This approach may be an alternative for analysis of BN, which requires sample turnarounds lower than five minutes and lower cost compared to traditional methods.

Application of Near-Infrared for Online Monitoring of Heavy Fuel Oil at Thermoelectric Power Plants. Part I: Development of Chemometric Models

“…The use of the whole spectrum can cause instability in the model, due to the noise variations presented in regions above 9000 cm –1 . Thus, models obtained with restricted wavelength simplify the interpretation allowing more accurate and robust predictions. , The selection of these regions was made by means of the “Optimize” tool available in the OPUS software and may vary depending on the property of interest. The Optimize tool suggests the best combination of spectral region and preprocessing to obtain the lowest RMSECV.…”