On the Validation Challenges of RANS-Based Transition Models

Luis Eça
Department of Mechanical Engineering, IST ULisboa, Portugal

In this presentation, we address mathematical modelling of engineering transitional flows and so we will focus on transition models for the Reynolds-averaged Navier-Stokes (RANS) equations. Furthermore, we will restrict ourselves to Local Correlation Transition Models (LCTM) that solve additional transport equations for transition dependent variables. This choice is justified by the present popularity of these models that are able to handle different types of transition onset and easy to include in most available RANS solvers. Nonetheless, some of the challenges discussed in this work are also valid for more sophisticated approaches as Large-Eddy Simulation (LES) or Direct Numerical Simulation (DNS). Simulating transitional flows with RANS based transition models generates additional challenges when compared to the traditional “fully-turbulent” simulations: effect of the problem definition including domain size and boundary conditions; numerical robustness and finally assessment of modelling accuracy (Validation). Furthermore, a given transition model may be coupled to different turbulence models with no guarantee that the different combinations produce similar results. On the other hand, best practice guidelines available in the open literature for the simulation of high Reynolds number flows with RANS may not be valid in the simulation of transitional flows. Last but not the least, assessment of the modelling accuracy requires a careful selection of quantities of interest and experimental data that might not be available. We discuss all these topics using three simple examples: the flow over a flat plate with natural and by-pass transition; the flow around an airfoil with separation-induced transition and the flow around a prolate spheroid that presents crossflow transition. Results obtained in these simple test cases show that the lack of complete experimental information hampers reliable assessment of modelling errors (Validation) of transition models. Nonetheless, there are two important trends found in the data: some of the improvements required may be related to the (background) turbulence model; present transition models improve RANS-based predictions of low Reynolds numbers flows when compared with the use of standard turbulence models.