Coarse-grid LES modeling based on machine-learning-based super-resolution reconstruction of under-resolved LES flows

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K18764
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 19:Fluid engineering, thermal engineering, and related fields
• Research Institution: Tohoku University
• Principal Investigator: Kawai Soshi 東北大学, 工学研究科, 教授 (40608816)
• Project Period (FY): 2022-06-30 – 2024-03-31
• Keywords: 数値流体力学 / 機械学習 / LES / 流体力学 / 乱流

Outline of Final Research Achievements

This study investigated a coarse-grid subgrid-scale (SGS) model that can maintain prediction accuracy as large-eddy simulation (LES) even when using a very coarse grid that intentionally does not resolve some high-energy turbulent components that should be resolved as LES. To establish the coarse-grid SGS model, we proposed a machine-learning-based pipeline model, which connects unsupervised and supervised learning, for super-resolution reconstruction of very coarse-grid under-resolved LES flowfields to provide effective SGS stresses for a coarse-grid LES. The proposed coarse-grid SGS model was validated through the apriori and a posteriori tests of turbulent channel flow.

Free Research Field

流体工学

Academic Significance and Societal Importance of the Research Achievements

高忠実に様々な乱流現象を再現可能とするLESは，近年は学術研究だけでなく，複雑な乱流現象を扱う必要がある産業界からの期待も非常に大きくなってきている．一方で，実際の設計開発におけるLESの活用は限定的なのもまた事実である．本研究がターゲットしているLESの飛躍的な低コスト化を目指した粗格子LESモデルの確立は，LESの利用を困難にしている高い計算コストの壁を取り除こうと試みるものであり，学術・応用の両面からLESの活用を大きく広げることに貢献することを目指して実施した．