Data-driven RANS Turbulence Modeling Approach for Hypersonic Vehicles

Corvid and NCSU, with the support of Leidos, will develop a machine learning-based framework to enhance RANS turbulence modeling at hypersonic conditions. We will start with the interrogation of existing LES databases from NCSU for model form identification and initial training, followed by a closed-loop iterative calibration procedure that will incorporate experimental data to further refine the framework. Initial evaluation of the modified RANS-based model will be performed on a set of validation cases and we will highlight the improvements made to surface flow predictions over standard RANS closure models, as well as the efficiency gains compared to high-fidelity methods. In later stages of the development, we will extend the framework to additional cases and demonstrate the improved accuracy and flexibility of the model for a wide range of conditions.