Construction of a three-component thermophysical property surrogate model for RP-3 aviation kerosene via machine learning-augmented molecular dynamics

Abstract

RP-3 aviation kerosene features a highly complex composition. The use of representative surrogate fuels offers a viable alternative for the accurate prediction of key thermophysical parameters. In this study, a ternary mixture of n-dodecane, n-propylcyclohexane, and n-propylbenzene was formulated to create a high-fidelity thermophysical property surrogate model for RP-3 aviation kerosene via machine learning-augmented molecular dynamics (MD-ML) simulations. MD simulation accurately predicts the thermophysical properties of pure components and their three-component models, showing close agreement with experimental data in terms of density, thermal conductivity, and viscosity. Through Gaussian process regression (GPR) optimization, an optimal surrogate formulation was identified, consisting of 60.65 wt% n-dodecane, 29.03 wt% n-propylcyclohexane, and 10.32 wt% n-propylbenzene. This approach enables precise prediction of thermophysical properties for RP-3 kerosene, providing important theoretical frameworks and technical tools for advanced aviation fuel design and combustion performance analysis.