• Corpus ID: 246035805

Learning physics-constrained subgrid-scale closures in the small-data regime for stable and accurate LES

  title={Learning physics-constrained subgrid-scale closures in the small-data regime for stable and accurate LES},
  author={Yifei Guan and Adam Subel and Ashesh Chattopadhyay and Pedram Hassanzadeh},
We demonstrate how incorporating physics constraints into convolutional neural networks (CNNs) enables learning subgrid-scale (SGS) closures for stable and accurate large-eddy simulations (LES) in the small-data regime (i.e., when the availability of high-quality training data is limited). Using several setups of forced 2D turbulence as the testbeds, we examine the a priori and a posteriori performance of three methods for incorporating physics: 1) data augmentation (DA), 2) CNN with group… 

Figures from this paper

Explaining the physics of transfer learning a data-driven subgrid-scale closure to a different turbulent flow
Novel analyses and a new framework are presented that explain the physics learned in TL and provide a framework to guide TL for wide-ranging applications in science and engineering, such as climate change modeling.


A priori analysis on deep learning of subgrid-scale parameterizations for Kraichnan turbulence
Different data-driven parameterizations for large eddy simulation of two-dimensional turbulence in the a priori settings are investigated and computational gain can be achieved using the intelligent eddy viscosity model that learns eddy Viscosity computed by the DSM instead of subgrid-scale stresses.
Data-driven subgrid-scale modeling of forced Burgers turbulence using deep learning with generalization to higher Reynolds numbers via transfer learning
It is shown that deep neural networks trained using properly pre-conditioned (augmented) data yield stable and accurate a posteriori LES models and that transfer learning enables accurate/stable generalization to a flow with 10x higher Reynolds number.
Investigation of nonlocal data-driven methods for subgrid-scale stress modeling in large eddy simulation
A nonlocal subgrid-scale stress (SGS) model is developed based on the convolution neural network (CNN), which is a powerful supervised data-driven method and also an ideal approach to naturally
Physical invariance in neural networks for subgrid-scale scalar flux modeling
From simulation-based experiments, it is shown that the proposed transformation-invariant NN model outperforms both purely data-driven ones as well as parametric state-of-the-art subgrid-scale models and is shown to generalize to regimes that have not been seen during the training phase.
Data-driven Equation Discovery of Ocean Mesoscale Closures
It is shown that efficient modern machine learning algorithms can accurately represent the physics of ocean eddies, can be constrained by physical laws, and can be interpretable and machine learning is now opening new avenues to improve climate simulations by extracting such parameterizations directly from data.
Applications of Deep Learning to Ocean Data Inference and Subgrid Parameterization
  • T. Bolton, L. Zanna
  • Environmental Science
    Journal of Advances in Modeling Earth Systems
  • 2019
The results indicate that data-driven approaches can be exploited to predict both subgrid and large-scale processes, while respecting physical principles, even when data are limited to a particular region or external forcing.
Subgrid modelling for two-dimensional turbulence using neural networks
The proposed methodology successfully establishes a map between inputs given by stencils of the vorticity and the streamfunction along with information from two well-known eddy-viscosity kernels, which represents a promising development in the formalization of a framework for generation of heuristic-free turbulence closures from data.
Investigations of data-driven closure for subgrid-scale stress in large-eddy simulation
Data-driven machine learning algorithms, random forests and artificial neural network (ANN), are used to establish the subgrid-scale (SGS) model for large-eddy simulation. A total of 30 flow