Prediction of FCC gasoline octane numbers using FT-MIR and PLS

A method for predicting octane numbers (RON and MON) in fluid catalytic cracking (FCC) gasolines is proposed. Using FT-MIR and PLS, improvements have been obtained in sample throughput, reduced delay times, accuracy (repeatability and reproducibility), amounts of samples and reagents and environmental working conditions when compared with current standard methods. A total number of 140 daily production samples were taken; and from there, a learning group was prepared (44 samples); a validation set (96 samples) was prepared, as well. Sample spectra were recorded from 4000 to 600 cm^-1 at 4 cm^-1 intervals (traditional sealed NaCl cells). The PLS technique was used in its two variants (1 and 2-block). Both provided similar results. Their predictive characteristics are very good: SEP_RON=0.38; SEP_MON=0.40; repeatability <0.1 O.N.; reproducibility <0.3 O.N. (SEP=Standard Error of Prediction).