Deep learning models turbulence closures for large eddy simulations

Researchers have developed a new deep learning approach for turbulence closure modeling in large eddy simulations (LES). This method uses a nudging technique, treating direct numerical simulation (DNS) data as sparse observations to train the model. This allows for a-priori training of closures, enabling the model to learn necessary forcing for accurate statistics while maintaining long-term stability without requiring backpropagation through the LES solver. AI

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

IMPACT Introduces a more stable and computationally efficient method for turbulence closure modeling in simulations.

RANK_REASON Academic paper detailing a novel deep learning approach for turbulence modeling.

Read on arXiv cs.LG →

• arXiv
• LES
• DNS

• paper
• other