A wall model for separated flows: embedded learning to improve a posteriori performance DOI Open Access
Zhideng Zhou, Xinlei Zhang, Guowei He

et al.

Journal of Fluid Mechanics, Journal Year: 2024, Volume and Issue: 1002

Published: Dec. 23, 2024

Developing large-eddy simulation (LES) wall models for separated flows is challenging. We propose to leverage the significance of flow data, which existing theories are not applicable, and knowledge wall-bounded (such as law wall) along with embedded learning address this issue. The proposed so-called features-embedded-learning (FEL) model comprises two submodels: one predicting shear stress another calculating eddy viscosity at first off-wall grid nodes. train former using wall-resolved LES (WRLES) data periodic hill wall. For latter, we a modified mixing length model, coefficient trained ensemble Kalman method. FEL assessed different configurations, resolutions Reynolds numbers. Overall good posteriori performance observed statistics recirculation bubble, stresses turbulence characteristics. modelled subgrid-scale (SGS) grids compared those calculated WRLES data. comparison shows that amplitude distribution SGS energy transfer obtained agree better reference when conventional model.

Language: Английский

Reynolds-Number-Dependence of Length Scales Governing Turbulent-Flow Separation in Wall-Modeled Large Eddy Simulation DOI
Rahul Agrawal, Sanjeeb Bose, Parviz Moin

et al.

AIAA Journal, Journal Year: 2024, Volume and Issue: 62(10), P. 3686 - 3699

Published: Aug. 23, 2024

This paper proposes a Reynolds number [Formula: see text] scaling for the of grid points required in wall-modeled Large Eddy Simulation (WMLES) turbulent boundary layers (TBL) to accurately capture regions flow separation. Based on various time scales nonequilibrium TBL, definition near-wall “underequilibrium” is proposed (in which “equilibrium” refers quasi balance between viscous and pressure gradient terms). length scale shown vary with as text]. A-priori analysis demonstrates that resolution ([Formula: text]) reasonably predict wall stress several flows at least text], irrespective Clauser parameter. Further, a-posteriori studies (on Boeing speed bump, Song– Eaton diffuser, Notre-Dame Ramp, backward-facing step) show such independent results accurate predictions separation same “nominal” across different numbers. Finally, we suggest near reattachment points, WMLES more restrictive than previous estimates by Choi Moin (Choi, H., Moin, P., “Grid-Point Requirements Simulation: Chapman’s Estimates Revisited,” Physics Fluids, Vol. 24, No. 1, 2012, Paper 011702) Yang Griffin (Yang, X. I. A., Griffin, K. Time-Step Direct Numerical Large-Eddy Simulation,” 33, 2021, 015108).

Language: Английский

Citations

4
Improved pressure-gradient sensor for the prediction of separation onset in RANS models DOI
Kevin P. Griffin, Ganesh Vijayakumar, Ashesh Sharma

et al.

Journal of Turbulence, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 18

Published: April 24, 2025

Language: Английский

Citations

0
Wall Modeling of Turbulent Flows with Varying Pressure Gradients Using Multi-Agent Reinforcement Learning DOI
Di Zhou, H. Jane Bae

AIAA Journal, Journal Year: 2024, Volume and Issue: 62(10), P. 3713 - 3727

Published: Aug. 5, 2024

We propose a framework for developing wall models large-eddy simulation that is able to capture pressure-gradient effects using multi-agent reinforcement learning. Within this framework, the distributed learning agents receive off-wall environmental states, including pressure gradient and turbulence strain rate, ensuring adaptability wide range of flows characterized by separations. Based on these determine an action adjust eddy viscosity and, consequently, wall-shear stress. The model training in situ with wall-modeled grid resolutions does not rely instantaneous velocity fields from high-fidelity simulations. Throughout training, compute rewards relative error estimated stress, which allows them refine optimal control policy minimizes prediction errors. Employing are trained two distinct subgrid-scale low-Reynolds-number flow over periodic hills. These validated through simulations hills at higher Reynolds numbers Boeing Gaussian bump. developed successfully acceleration deceleration wall-bounded turbulent under gradients outperform equilibrium predicting skin friction.

Language: Английский

Citations

1
A wall model for separated flows: embedded learning to improve a posteriori performance DOI Open Access
Zhideng Zhou, Xinlei Zhang, Guowei He

et al.

Journal of Fluid Mechanics, Journal Year: 2024, Volume and Issue: 1002

Published: Dec. 23, 2024

Developing large-eddy simulation (LES) wall models for separated flows is challenging. We propose to leverage the significance of flow data, which existing theories are not applicable, and knowledge wall-bounded (such as law wall) along with embedded learning address this issue. The proposed so-called features-embedded-learning (FEL) model comprises two submodels: one predicting shear stress another calculating eddy viscosity at first off-wall grid nodes. train former using wall-resolved LES (WRLES) data periodic hill wall. For latter, we a modified mixing length model, coefficient trained ensemble Kalman method. FEL assessed different configurations, resolutions Reynolds numbers. Overall good posteriori performance observed statistics recirculation bubble, stresses turbulence characteristics. modelled subgrid-scale (SGS) grids compared those calculated WRLES data. comparison shows that amplitude distribution SGS energy transfer obtained agree better reference when conventional model.

Language: Английский

Citations

1