Abstract:
Sustainable aviation fuels (SAFs) are essential to decarbonizing long-haul aviation, but their deployment is constrained by slow, data-intensive certification protocols and limited understanding of how novel molecules and blends behave in real combustors. This dissertation develops an integrated framework that couples data-efficient artificial intelligence models with high-fidelity computational fluid dynamics (CFD) simulations of spray ignition to accelerate SAF development. First, we introduce Adaptive Checkpointing with Specialization (ACS), a new training scheme for multi-task graph neural networks operating in ultra-low-data regimes. Across benchmark molecular datasets and a SAF case study spanning 1,379 molecules and 15 properties, ACS matches or exceeds state-of-the-art baselines while achieving reliable accuracy with as few as 29 labeled samples, and delivers >20% gains over conventional training schemes. Second, we develop a framework for validating reduced chemical kinetic mechanisms using spray-ignition simulations in constant-volume chambers. This approach enables the assessment of low-volatility jet-fuel components that cannot be tested in conventional fundamental devices. By resolving mixture stratification, vaporization cooling, and flame-kernel formation, the model reproduces experimentally observed ignition behavior and demonstrates that a 1/7-sector sectorization can reduce CPU cost by ~50% while maintaining predictive accuracy. Together, these contributions deliver a coherent AI–CFD pipeline that screens SAF candidates, predicts their properties across molecules and mixtures, and validates their ignition behavior under realistic spray conditions—offering a practical path to faster, more informed SAF certification and design.

Biography:
Basem Abdulmuhsen Eraqi is a Ph.D. candidate in Mechanical Engineering program in the PSE division at KAUST, supervised by Prof. Mani Sarathy. He received his Bachelor of Science degree in Mechanical Engineering from the University of California, Irvine, in 2021, followed by a Master of Science degree in Mechanical Engineering from KAUST in 2022. He joined the Combustion and Pyrolysis Chemistry (CPC) research group at KAUST as a Master's student in 2021. His research focuses on artificial intelligence–driven modeling of sustainable aviation fuels, with an emphasis on molecular property prediction, data-efficient machine learning frameworks, and the integration of chemical knowledge with graph neural networks. His work spans multi-task learning, ultra-low-data regimes, and the application of AI to fuel design and performance assessment, supported by experimental and high-fidelity modeling insights.