Robust deep learning for emulating turbulent viscosities

TL;DR

This paper introduces a robust patch-based deep learning approach to accurately emulate turbulent viscosities in fluid flows, demonstrating improved efficiency and accuracy over traditional methods using the Spallart-Allmaras model.

Contribution

The study presents a novel patch-based training strategy for deep neural networks to model turbulence, enhancing accuracy and reducing computational costs.

Findings

Patch-based training improves accuracy in turbulence modeling.

The method reduces training computational costs.

Effective on the Spallart-Allmaras turbulence model.

Abstract

From the simplest models to complex deep neural networks, modeling turbulence with machine learning techniques still offers multiple challenges. In this context, the present contribution proposes a robust strategy using patch-based training to learn turbulent viscosity from flow velocities, and demonstrates its efficient use on the Spallart-Allmaras turbulence model. Training datasets are generated for flow past two-dimensional (2D) obstacles at high Reynolds numbers and used to train an auto-encoder type convolutional neural network with local patch inputs. Compared to a standard training technique, patch-based learning not only yields increased accuracy but also reduces the computational cost required for training.