Dynamic Deep Learning LES Closures: Online Optimization With Embedded DNS

TL;DR

This paper introduces a novel online training method for deep learning-based LES closure models that dynamically adapts during simulations using embedded DNS data, improving accuracy across different geometries and regimes.

Contribution

It presents an innovative online optimization approach for training deep learning LES closures during simulations with embedded DNS, addressing data scarcity and overfitting issues.

Findings

Closure models trained online adapt to specific geometries.

Improved accuracy over traditional offline-trained models.

Demonstrated effectiveness in turbulent flow simulations.

Abstract

Deep learning (DL) has recently emerged as a candidate for closure modeling of large-eddy simulation (LES) of turbulent flows. High-fidelity training data is typically limited: it is computationally costly (or even impossible) to numerically generate at high Reynolds numbers, while experimental data is also expensive to produce and might only include sparse/aggregate flow measurements. Thus, only a relatively small number of geometries and physical regimes will realistically be included in any training dataset. Limited data can lead to overfitting and therefore inaccurate predictions for geometries and physical regimes that are different from the training cases. We develop a new online training method for deep learning closure models in LES which seeks to address this challenge. The deep learning closure model is dynamically trained during a large-eddy simulation (LES) calculation using embedded direct numerical simulation (DNS) data. That is, in a small subset of the domain, the flow is computed at DNS resolutions in concert with the LES prediction. The closure model then adjusts its approximation to the unclosed terms using data from the embedded DNS. Consequently, the closure model is trained on data from the exact geometry/physical regime of the prediction at hand. An online optimization algorithm is developed to dynamically train the deep learning closure model in the coupled, LES-embedded DNS calculation.