No abstract

The ability of near-infrared (NIR) spectroscopy with multivariate methods of analysis to predict the chemical composition of different process streams obtained from a catalytic reforming unit was demonstrated. One hundred and forty samples were used to develop calibration models by partial least-squares (PLS) regression. For calibration and validation stages, 92 and 48 samples were employed, respectively. Total paraffin, total isoparaffin, total naphthene, and total aromatic content in naphtha have been successfully determined. Prediction of individual carbon chain length (C6-C8) in each hydrocarbon family was also studied with promising results. Calibration models for compounds with carbon chain length smaller than C6 and bigger than C8 were not developed because the concentration ranges of those components were too low, and there is a limitation in NIR sensitivity. The proposed methodology takes less than 5 min to perform, and it can be used for online process control. Besides, it is faster than the standard method, which takes about 4 h. The results showed a high repeatability and a good correlation with the GC data.

No abstract

In this paper a comparison is made between the performance of models developed by applying chemometric analysis to NIR and UVVIS spectral data obtained from feedsctock samples corresponding to the different Ecopetrol S.A., Barrancabermeja Refinery FCC units for predicting some important physicochemical properties. The results show the utility of both methodologies here evaluated to follow up the quality of these types of refinery streams and present the advantages and disadvantages of each methodology for predicting the feedstock properties here evaluated.