NN-augmented k - ω shear stress transport turbulence model for high-speed flows with shock-wave/boundary layer interaction

Menter k-ω shear stress transport (SST) turbulence model demonstrates excellent performance for incompressible, subsonic and transonic flows with mild separation but shows overprediction of the separation bubble of supersonic shock-wave/boundary layer interaction (SWBLI). Some efforts focus on the effect of the structure parameter in stress limiter in an ad-hoc way. Few studies attempt to construct the relation between the structure parameter and flow field variables. The motivation of this work is to construct such a relation to augment the prediction performance of the SST model by introducing a correction factor. Machine learning methods are used since the physical mechanism of SWBLI is complex and unclear. The simulation results show that the constructed relation enhances the structure parameter near the shock wave in the boundary layer when applied to the SST model. Compared with direct numerical simulation and experimental data, the pressure and skin friction coefficients along the wall and the velocity field are significantly improved. In addition, the introduced correction factor can automatically degrade for the subsonic benchmark case of NACA4412 airfoil and maintain the prediction accuracy of the original SST model, but delay the shock location of the transonic case.