A GPU-Enabled Building Block Flow Model for Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) is an key tool in the design of aircraft, allowing engineers to predict the performance of a configuration without having to conduct expensive physical tests. However, in order to move to a greater reliance on CFD, the industry requires a high level of accuracy and fast turnaround time, which current methods cannot deliver. In recent years, the rapid development of the GPU industry has led to an explosion of computational power with the GPU architecture. This has allowed wall-modeled large eddy simulation (WMLES), a higher fidelity simulation technique, to become practical for industry use. WMLES requires the use of both a sub-grid scale (SGS) model and a wall model in order to close the system of equations for integration. Although WMLES delivers an improvement over previous methods, classical SGS and wall models do not deliver the accuracy required by the aviation industry. To help close this gap, we introduce a GPU-compatible version of the Building-Block Flow Model (BFM), a machine learning based unified sub-grid scale and wall model for LES introduced in [1]. In this thesis, we discuss the implementation of the BFM for GPU, timing of the BFM versus other closure models for WMLES, and a variety of tests with the BFM designed to evaluate its performance, and possible avenues of improvement.

Name thesis:degree_name

Master

Department dc:contributor.department

Massachusetts Institute of Technology. Department of Aeronautics and Astronautics

Grantor dc:publisher

Massachusetts Institute of Technology

Year dc:date.issued

2024

Author dc:creator

• Costa, Samuel Thomas

Advisor dc:contributor.advisor

• Lozano-Durán, Adrián

Statement dc:rights

• Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
• Copyright retained by author(s)

Licence dc:rights.uri

• https://creativecommons.org/licenses/by-nc-nd/4.0/

Handle dc:identifier.uri

https://hdl.handle.net/1721.1/157810

OAI identifier oai:identifier

oai:dspace.mit.edu:1721.1/157810