Neural Network for Subgrid Turbulence Modeling for Large Eddy Simulations

Eduardo Vital, Jean-Marc Gratien, Yassine Ayoun, Thibault Faney, Julien Bohbot

arXiv:2511.05103·physics.flu-dyn·Published 2025-11-07·Updated 2025-11-27

When simulating multiscale systems, where some fields cannot be fully prescribed despite their effects on the simulation's accuracy, closure models are needed. This phenomenon is observed in turbulent fluid dynamics, where Large Eddy Simulations (LES) depict global behavior while turbulence modeling introduces dissipation correspondent to smaller sub-grid scales. Recently, scientific machine learning techniques have emerged to address this problem by integrating traditional (physics-based) equations with data-driven (machine-learned) models, typically coupling numerical solvers with neural networks. This work presents a comprehensive workflow, encompassing high-fidelity data generation and post-processing, a priori learning, and a posteriori testing, where data-driven models enrich differential equations.

TopicsScientific Machine Learning & PINNs

Tagsles scientific-machine-learning turbulence-modeling

arXiv categoriesphysics.flu-dyn, physics.comp-ph

arXiv abstract pagePDF