Machine learning parametrizes subgrid scales in fluid dynamics simulations

Researchers have developed a novel method for simulating fluid dynamics by employing a feed-forward neural network to learn sub-grid fluxes within the Shallow Water equations. This approach utilizes a local parametrization with a four-point stencil, offering advantages over globally coupled methods. The machine learning technique has demonstrated improved energy balance in long-term turbulent simulations and can be combined with flux limiting for enhanced accuracy near shocks, even in untrained dynamical regimes. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel ML approach for fluid dynamics simulations, potentially improving accuracy and efficiency in long-term turbulent modeling.

RANK_REASON Academic paper detailing a new machine learning method for fluid dynamics simulations. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Ilya Timofeyev
• Shallow Water equations
• arXiv
• Hugging Face

• paper
• other