Document Type

Conference Paper

Publication Date

2023

Conference Name

AIAA SciTech Forum, January 23-27, 2023, National Harbor, MD

Abstract

This paper expands the authors’ prior work[1], which focuses on developing a convolutional neural network (CNN) model capable of mapping time-averaged, unsteady Reynold’s-averaged Navier-Stokes (URANS) simulations to higher resolution results informed by time-averaged detached eddy simulations (DES). The authors present improvements over the prior CNN autoencoder model that result from hyperparameter optimization, increased data set augmentation through the adoption of a patch-wise training approach, and the predictions of primitive variables rather than vorticity magnitude. The training of the CNN model developed in this study uses the same URANS and DES simulations of a transonic flow around several NACA 4-digit airfoils at high angles of attack[1]. The authors test the updated model by inputting airfoil profiles and flow conditions outside of the training set and by comparing the output flow field against DES calculations. The Fully Unstructured Navier-Stokes 3D (FUN3D) solver from NACA generates the computational fluid dynamics (CFD) simulations and uses computing assets available from the Department of Defense High Performance Computing Modernization Program (HPCMP) and Old Dominion University (ODU) High Performance Computing (HPC). Finally, the paper includes the effects of these techniques on the predictive capability and the performance of the authors’ CNN model.

Rights

This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.

Original Publication Citation

Romano, J., Brodeur, A. C., & Baysal, O. (2023). Patch-wise training to improve convolutional neural network synthetic upscaling of computational fluid dynamics simulations. [Paper presentation]. AIAA SciTech Forum, National Harbor, MD. https://doi.org/10.2514/6.2023-1804

ORCID

0000-0002-7041-3869 (Baysal)

Share

COinS