Machine Learning-Based Molecular Dynamics Studies on Predicting Thermophysical Properties of Ethanol–Octane Blends

This paper presents an innovative approach to predicting thermophysical properties of ethanol–octane blends by integrating molecular dynamics (MD) simulations with machine learning (ML) algorithms. The work addresses the growing interest in ethanol–gasoline blends as alternative fuels and the need for efficient computational methods to analyze their properties. Using MD simulations and various ML models such as Decision Tree Regression (DTR), Random Forest Regression (RFR) and Gaussian Process Regression (GPR), the behavior of 660-molecule systems of ethanol–octane mixtures was modeled. The OPLS-AA force field was employed to accurately represent the molecular interactions. Among the ML models, DTR demonstrated the highest accuracy in predicting atomic displacements and velocities. The integration of MD with ML promises rapid and accurate predictions, with error rates consistently below 2.5% across different ethanol concentrations and timesteps. Notably, the ML model showcases remarkable speedup in computational efforts, approximately 1.8, 2.7, and 3.4 times faster for E10, E20 and E85 blends respectively compared with the traditional MD simulations. This approach not only enhances the understanding of ethanol–octane blend properties but also demonstrates the potential for ML to accelerate the complex molecular simulations. The findings of this study have significant implications for the design and optimization of alternative fuels, targeting the sustainable energy demand.