Advancing Sustainable Aviation Fuel Design: Machine Learning for High-Energy-Density Liquid Polycyclic Hydrocarbons

This study explores the identification of polycyclic hydrocarbons (PCHCs) with high energy density (HED) using machine learning (ML) techniques, specifically focusing on establishing a quantitative structure-property relationship (QSPR). The support vector machine (SVM) algorithm was employed for its strong predictive performance of net heat of combustion (NHOC), achieving a high coefficient of determination (R²) and a low mean absolute error (MAE) of 27.821 kJ/mol for 20% test data using only six key descriptors. From reputable scientific literature and databases, 35 potential HED PCHCs (ranging from C6 to C15) were identified. Structural analysis showed that these PCHCs predominantly consist of saturated alkanes featuring multiple triangular, rectangular, and pentagonal rings, highlighting the significant role of strain energy in achieving HED. This study emphasizes the importance of specific energy and energy density as primary considerations for sustainable aviation fuel (SAF) design, while also recognizing the need to meet additional properties and comply with ASTM D7566/D4054 standards. This work successfully achieves the initial objectives of our SAF program, laying a robust foundation for the further development of high-performance, sustainable aviation fuels.