Kinematic viscosity prediction of jet fuels and alternative blending components via comprehensive two-dimensional gas chromatography, partial least squares, and Yeo-Johnson transformation.

This work presents an accurate yet simplified partial least squares model to predict the kinematic viscosity of conventional and alternative jet fuels at -20°C using comprehensive two-dimensional gas chromatography coupled to a flame ionization detector (GC × GC/FID). Three different normalization methods (mean-centering, logarithmic, and Yeo-Johnson) were evaluated to identify their impact in the prediction of middle distillates' physical properties. Results using Yeo-Johnson transformation exhibited improved viscosity prediction capabilities over the validation set with a mean absolute percentage error of 5.3%, a root-mean-squared error of 0.23, and a coefficient of determination (R ) of 0.9404 using only 10 latent variables. Unlike previously reported correlations, this model allowed the identification of specific hydrocarbon groups and carbon numbers that drive jet fuel viscosity at low temperatures. The presence of even small amounts of large branched-alkanes (C -C ), dicyclic-alkanes (C ), and cycloaromatics (C ) have the potential to strongly increase the kinematic viscosity of jet fuels. Contrastingly, light monocycloalkanes and branched-alkanes (≤ C ) were associated with lower viscosity values. Novelly, this model suggests the implementation of Yeo-Johnson transformations to predict the physical properties of middle distillates to further improve the performance metrics of partial least squares models based on GC data.