Improved prediction of fuel properties with near-infrared spectroscopy using a complementary sequential fusion of scatter correction techniques

Near-infrared (NIR) spectroscopy of fuels can suffer from scattering effects which may mask the signals corresponding to key analytes in spectra. Therefore, scatter correction techniques are often used prior any modelling so remove and improve predictive performances. However, different carry complementary information that, if jointly used, both model stability performances could be improved. A solution that is fusion differently corrected data. In present work, use a preprocessing approach called sequential through orthogonalization (SPORT) demonstrated for predicting quality parameters diesel fuels. particular, possibility four properties, i.e., boiling point (°C), density (g/mL), aromatic mass (%) viscosity (cSt), was considered. As comparison, standard partial least-squares (PLS) regression performed on data pretreated by SNV 2nd derivative (which widely combination). The results showed SPORT models, based techniques, outperformed PLS models prediction all suggesting selection single technique not sufficient. Up complete bias removal with 50% reduction error obtained. R2P increased up 8%. limited NIR but applied other spectral where optimization step required.