Molecular dynamics simulations coupled with machine learning for investigating thermophysical properties of binary surrogate aviation kerosene

Abstract

During the scramjet regenerative cooling, aviation kerosene in a supercritical state exhibits thermophysical properties that differ significantly from those at ambient conditions. Since n-decane and n-propylcyclohexane are commonly used as surrogate fuels, understanding their thermophysical properties is essential for practical applications. This study employs equilibrium state molecular dynamics (EMD) simulations with TraPPE-UA and OPLS force fields to investigate key properties of density, thermal conductivity and viscosity for both pure component and their mixtures. The results indicate that TraPPE-UA accurately simulates density and thermal conductivity, while OPLS provides superior viscosity predictions. Machine learning techniques based on MD simulation outcomes demonstrate a strong ability to predict thermophysical properties, significantly reducing simulation time. The study further examines temperature-driven molecular structural changes, revealing that molecular distortions hinder heat transfer, while increased molecular spacing facilitates flow and reduces viscosity. This work provides an effective approach for studying the thermophysical properties of aviation kerosene.