03.1 - Aerodynamics – CFD Methods and Validation
DATA-DRIVEN TURBULENCE MODELING FOR MULTI-MECHANISM SEPARATED FLOWS OVER ICED AIRFOILS
Z.Z. He¹, Y.F. Zhang¹; ¹Tsinghua University, China
Accurate iced wing aerodynamic prediction is hindered by the "dual error mechanism" in RANS models, where eddy viscosity is overestimated upstream and underestimated downstream in separated shear layers. This study proposes the field inversion and physics-driven-classification modeling (FI-PDCM) framework to decouple these opposing error modes. A correction factor 𝛽 is first determined via discrete adjoint-based field inversion, followed by the formulation of an explicit algebraic model using local features. Validation on the GLC305-944 iced airfoil benchmark demonstrates that the corrected model accurately replicates the 𝛽 distribution derived from field inversion, capturing the separation bubble physics, surface pressure, and the entire lift curve through stall. Furthermore, tests on a zero-pressure-gradient flat-plate flow confirm that the model effectively deactivates in attached regions, preserving the original predictive accuracy for canonical wall-bounded flows. Future research will focus on testing the model’s generalizability across a broader range of ice shapes to further establish its robustness and engineering utility.