Deep Learning Methods for Reynolds-Averaged Navier-Stokes Simulations of Airfoil Flows

TL;DR

This paper evaluates deep learning models, specifically a modernized U-net, for accurately predicting RANS solutions for airfoil flows, demonstrating less than 3% error and emphasizing the influence of training data and model size.

Contribution

It introduces a comprehensive analysis of deep learning for RANS simulations, highlighting the impact of data and model complexity on accuracy, and provides publicly available code for reproducibility.

Findings

Achieved less than 3% mean relative error in pressure and velocity predictions.

Training data size and network weights significantly affect model accuracy.

Source code is publicly available for further research and validation.

Abstract

With this study we investigate the accuracy of deep learning models for the inference of Reynolds-Averaged Navier-Stokes solutions. We focus on a modernized U-net architecture, and evaluate a large number of trained neural networks with respect to their accuracy for the calculation of pressure and velocity distributions. In particular, we illustrate how training data size and the number of weights influence the accuracy of the solutions. With our best models we arrive at a mean relative pressure and velocity error of less than 3% across a range of previously unseen airfoil shapes. In addition all source code is publicly available in order to ensure reproducibility and to provide a starting point for researchers interested in deep learning methods for physics problems. While this work focuses on RANS solutions, the neural network architecture and learning setup are very generic, and applicable to a wide range of PDE boundary value problems on Cartesian grids.