摘要: We perform wall-modeled large-eddy simulation (WMLES) of the Gaussian bump using the latest implementation of the Building-block Flow model (BFM) [1, 2]. BFM is a unified subgrid-scale (SGS) and wall model for WMLES that devises t... 展开 We perform wall-modeled large-eddy simulation (WMLES) of the Gaussian bump using the latest implementation of the Building-block Flow model (BFM) [1, 2]. BFM is a unified subgrid-scale (SGS) and wall model for WMLES that devises the flow as a collection of building blocks. The model is rooted in the assumption that simple canonical flows contain the essential physics to provide accurate predictions in more complex scenarios [3]. BFM is implemented via artificial neural networks and accounts for a limited set of wall-attached turbulence, adverse pressure gradient effects, and separation. The first version of BFM has been already tested on a complex aircraft geometry, showing improvements with respect to standard WMLES approaches [1]. Here, we assess the robustness and capabilities of the second version of BFM in the Gaussian Bump benchmark with emphasis on affordable grid resolutions. The results are compared with available experimental data, as well as results computed using standard WMLES. BFM captures the location and size of the separation bubble with less than 1% error for the coarse grid resolutions considered, improving on standard WMLES approaches. The error in the prediction of the pressure coefficient, friction coefficient, and mean velocities profiles in the second half of the separation bubble is also improved and robust under different grid resolutions. However, the results are still unsatisfactory in the first half of the separation bubble. The mismatch was explained by the lack of training data for BFM to account for strong flow reversal. Future versions of BFM will incorporate additional adverse pressure gradient effects to enhance model performance. 收起