Influence of Branching Degree-Derived <i>iso</i>-Alkane GC×GC Subgroups on Fuel Property Prediction

This study addresses a fundamental source of uncertainty in predicting the properties of sustainable aviation fuels (SAF) based on their composition: the unresolved distribution of structural isomers within iso-alkanes. Two complementary weighted average models are applied to analyze the influence of subgroups in composition analysis on the prediction of fuel properties. Both models are based exclusively on isomer properties and mixing rules without any training or fitting process. The first, the Mean Matrix model, represents fuels using conventional two-dimensional gas chromatography (GC×GC) resolution (carbon number by hydrocarbon family), whereas the second, the SubGroup Mean Matrix (SGMM) model, incorporates retention index (RI) based subgroups within the C7–C17 iso-alkane range to capture structural differences among isomers. Both models are built from a comprehensive isomer database augmented by Quantitative Structure–Property Relationship (QSPR) predictions where experimental data was unavailable. The model performance is evaluated on eight samples (Jet A-1; ATJ-SPK, two FT-SPK and four paraffinic solvents), comparing absolute errors and a normalized Relative Performance Change (RPC) metric. By incorporating higher compositional resolution via iso-alkane subgroups in the SGMM, the predictive accuracy for all evaluated properties of the ATJ-SPK fuel, which is characterized by a narrow and highly branched composition, was significantly improved. These properties included distillation temperatures, density, viscosity, net heat of combustion, and cetane number. In general, the improvements in volatile properties are most noticeable. Bulk properties such as density and net heat of combustion show only minor changes in prediction accuracy. Although the approach improves or maintains the predictive accuracy for the fuel samples, a deterioration can also be observed to some extent in the property prediction of the paraffinic solvents. Overall, the results indicate that incorporating subgroup-level resolution not only improves prediction accuracy for fuels with narrow, highly branched isomer profiles, but also can lead to an average performance improvement for most of the samples.