Towards a Robust, Generalizable RANS Model for Heat Transfer Predictions in Hypersonic Boundary Layers

Traditional RANS turbulence models fail to accurately predict heat transfer within hypersonic boundary layers, especially in the presence of shock interactions. While there have been attempts to improve this by incorporating more accurate physics or using machine learning techniques, many of the resulting models are either restricted to a narrow set of flows; fail to offer improvements or reliable results when extrapolated to new, but related flows; or are difficult to implement and reproduce by others. This work details the progress made in developing a robust, generalizable, data-driven augmentation for improving heat transfer predictions within hypersonic boundary layers for RANS solvers. The effects of improved models for anisotropy, the turbulent Prandtl number, and non-equilibrium effects are investigated using inverse modeling approaches and translated into requirements for an ideal RANS model. This knowledge is used to propose and learn an augmentation to the Wilcox-2006 k-ω turbulence model. The augmentation frameworks used and procedures followed are discussed, and preliminary results are presented, with careful consideration given to reproducibility.