Probabilistic Mean Quantitative Structure–Property Relationship Modeling of Jet Fuel Properties

We present a novel probabilistic mean quantitative structure–property relationship (M-QSPR) method for the prediction of jet fuel properties considering two-dimensional gas chromatography measurements. Fuels are represented as one pseudo-structure that is inferred by weighted average over structures 1866 molecules could be in individual fuel. The allows training models on both data pure components and fuels does not require mixing rules calculation bulk property. This drastically increases number available direct learning behavior. For modeling, we use Monte-Carlo dropout neural network, machine algorithm, estimates uncertainties due to possible unidentified isomers dissimilarity test data. Models developed predict freezing point, flash net heat combustion, temperature-dependent such density, viscosity, surface tension. investigate effect presence predictions up 82 conventional 50 synthetic fuels. results compared three metrics quantify accuracy, precision, reliability. These allow comprehensive estimation predictive capability models. tension, M-QSPR yields highly accurate even without with nonlinear behavior temperature complex component interactions, like viscosity was found essential method.