Exploration of robust machine learning strategy for subgrid scale stress modeling

被引:4
|
作者
Abekawa, A. [1 ]
Minamoto, Y. [1 ]
Osawa, K. [2 ]
Shimamoto, H. [1 ]
Tanahashi, M. [1 ]
机构
[1] Tokyo Inst Technol, Sch Engn, Ookayama, Meguro, Tokyo 1528550, Japan
[2] Univ Politecn Madrid, Sch Aeronaut, Madrid 28040, Spain
关键词
TURBULENT; FLOW; SIMULATION; EDDIES;
D O I
10.1063/5.0134471
中图分类号
O3 [力学];
学科分类号
08 ; 0801 ;
摘要
Various aspects of machine learning (ML) are explored to resolve limitations appearing in current ML-based subgrid scale (SGS) stress modeling. Graph neural network (GNN), applied in the present study, allows flexible and rigorous use of spatial convolution regardless of the proximity to physical boundaries and mesh uniformity. Along with GNN, the proposed feature scaling method relies only on the local quantities and can be applied for a range of flow configurations. A data augmentation method is also proposed to consider the rotational invariant. All these techniques are implemented in the present model, and the model is compared with versions of corresponding ML-based models including a typical multilayer perceptron (MLP) for various flow configurations. The results showed that both GNN and MLP models yield reasonable prediction overall. However, GNN shows superior performance near-wall due to spatial convolution. Although the present method implements the rotational invariant discretely, the augmentation method is found to produce consistent performance for any rotated coordinates. The minimal flow configuration, which can train a model to predict a range of flow configurations, is also explored. It is found that a model trained based on turbulent channel flows alone yields a close level of prediction robustness to the ones trained with multiple flow configurations. The developed GNN model is implemented in OpenFOAM, and large eddy simulation (LES) results are compared with corresponding direct numerical simulation data. With these proposed techniques, ML-based SGS models can be improved in terms of robustness and usability for a range of LES applications.
引用
收藏
页数:16
相关论文
共 50 条
  • [41] Probabilistic Regularized Extreme Learning Machine for Robust Modeling of Noise Data
    Lu, XinJiang
    Ming, Li
    Liu, WenBo
    Li, Han-Xiong
    IEEE TRANSACTIONS ON CYBERNETICS, 2018, 48 (08) : 2368 - 2377
  • [42] EDDY DIFFUSIVITY, EDDY NOISE AND SUBGRID-SCALE MODELING
    ROSE, HA
    JOURNAL OF FLUID MECHANICS, 1977, 81 (AUG5) : 719 - 734
  • [43] Modeling subgrid-scale effects on particles by approximate deconvolution
    Shotorban, B
    Mashayek, F
    PHYSICS OF FLUIDS, 2005, 17 (08) : 1 - 4
  • [44] Benchmarking of Machine Learning Ocean Subgrid Parameterizations in an Idealized Model
    Ross, Andrew
    Li, Ziwei
    Perezhogin, Pavel
    Fernandez-Granda, Carlos
    Zanna, Laure
    JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS, 2023, 15 (01)
  • [45] Implicit subgrid-scale modeling by adaptive local deconvolution
    Adams, N. A.
    Direct and Large-Eddy Simulation V, Proceedings, 2004, 9 : 15 - 22
  • [46] Subgrid Scale Combustion Modeling Based on Stochastic Model Parameterization
    Calhoon, William H., Jr.
    Zambon, Andrea C.
    Sekar, Balu
    Kiel, Barry
    JOURNAL OF ENGINEERING FOR GAS TURBINES AND POWER-TRANSACTIONS OF THE ASME, 2012, 134 (03):
  • [47] Hyperviscosity and Vorticity-Based Models for Subgrid Scale Modeling
    G. Dantinne
    H. Jeanmart
    G.S. Winckelmans
    V. Legat
    D. Carati
    Applied Scientific Research, 1997, 59 (4) : 409 - 420
  • [48] Filtered density function for subgrid scale modeling of turbulent combustion
    Givi, Peyman
    AIAA Journal, 2006, 44 (01): : 16 - 23
  • [49] Hyperviscosity and vorticity-based models for subgrid scale modeling
    Dantinne, G
    Jeanmart, H
    Winckelmans, GS
    Legat, V
    Carati, D
    APPLIED SCIENTIFIC RESEARCH, 1998, 59 (04): : 409 - 420
  • [50] Subgrid scale modeling of droplet bag breakup in VOF simulations
    Han, Austin
    Desjardins, Olivier
    INTERNATIONAL JOURNAL OF MULTIPHASE FLOW, 2024, 180