Implementation of a TBLE-Based Wall Model With Pressure Gradient in a Massively Parallel LES Solver
Résumé
Abstract Wall modeling is a common approach to reduce the cost of Large-Eddy Simulations (LES) of high Reynolds number flows. The selection of the wall model is however crucial due to the underlying assumptions made to derive these models. The most frequently employed wall models rely on strong hypotheses such as the absence of pressure gradient, curvature, and the assumption of stationarity. These assumptions are questionable for turbomachinery configurations, which can involve complex phenomena and may result in mispredictions. This work aims to investigate the resolution of a Thin Boundary Layer Equation (TBLE) approach including the pressure gradient effect and evaluate the gain compared to a classical logarithmic law approach. The proposed approach accounts for the pressure gradient directly through a source term. Two configurations are investigated to validate this approach and its performance. First, a planar turbulent channel without pressure gradient is investigated to ensure the model’s ability to accurately predict a turbulent boundary layer. The second case, the periodic hill test case available on the ERCOFTAC database, is used to assess the wall model’s performance on a flow submitted to pressure gradient effects. The conclusions of this work are twofold. First, such wall models can be sensitive to the unsteady pressure gradient fluctuations present in the LES prediction even for cases without mean pressure gradient; temporal averaging can thus be required to reach satisfying results. Second, substantial improvement is found in the periodic hill problem if compared to the algebraic wall model, in terms of mean velocity profiles, recirculation bubble size, reattachment zone, and friction profiles. Promising contributions of such tools can therefore be expected for turbomachinery configurations.