Recurrent multi-mesh convolutional autoencoder framework for spatio-temporal aerodynamic modelling
Recurrent multi-mesh convolutional autoencoder framework for spatio-temporal aerodynamic modelling
This work presents a geometric-deep-learning multi-mesh autoencoder framework for static and unsteady aerodynamic predictions. Aerodynamic analyses on real-world aerospace applications, such as flow around an aircraft, using high-fidelity solvers is computationally prohibitive, especially in unsteady and free-moving bodies. To solve the limitation of the computational burden, reduced order models capable of replicating aerodynamic responses at a fraction of the computing cost result adequate. Deep learning presents a suite of new methods capable of inferring more complex representations from data involving large and highly nonlinear systems. Thus, such data-driven technology is particularly suited for aerodynamic simulations. The proposed framework leverages on geometric deep learning for aerodynamic surface meshes embedded with a multi-resolution algorithm for dimensionality reduction and a recurrent scheme for time-integrated solutions. A building-block implementation allows, with the same framework, to predict superficial aerodynamic quantities in distinct conditions: steady, unsteady or dynamic; and in highly nonlinear transonic regimes. The developed methodology is validated in relevant test cases, proving the applicability of our framework to different problems. The computational efficiency against traditional high-fidelity solvers is demonstrated, obtaining computational gains of well over 99.9% on new simulations. Furthermore, a comprehensive comparison between a quasi-steady and a recurrent scheme for forced-motion responses is presented.
Aerospace Research Central
Massegur Sampietro, David
d5bc71e8-f1b8-4c9f-9537-7ff63ad19426
Da Ronch, Andrea
a2f36b97-b881-44e9-8a78-dd76fdf82f1a
Massegur Sampietro, David
d5bc71e8-f1b8-4c9f-9537-7ff63ad19426
Da Ronch, Andrea
a2f36b97-b881-44e9-8a78-dd76fdf82f1a
Massegur Sampietro, David and Da Ronch, Andrea
(2023)
Recurrent multi-mesh convolutional autoencoder framework for spatio-temporal aerodynamic modelling.
In AIAA AVIATION 2023 Forum.
Aerospace Research Central..
(doi:10.2514/6.2023-3845).
Record type:
Conference or Workshop Item
(Paper)
Abstract
This work presents a geometric-deep-learning multi-mesh autoencoder framework for static and unsteady aerodynamic predictions. Aerodynamic analyses on real-world aerospace applications, such as flow around an aircraft, using high-fidelity solvers is computationally prohibitive, especially in unsteady and free-moving bodies. To solve the limitation of the computational burden, reduced order models capable of replicating aerodynamic responses at a fraction of the computing cost result adequate. Deep learning presents a suite of new methods capable of inferring more complex representations from data involving large and highly nonlinear systems. Thus, such data-driven technology is particularly suited for aerodynamic simulations. The proposed framework leverages on geometric deep learning for aerodynamic surface meshes embedded with a multi-resolution algorithm for dimensionality reduction and a recurrent scheme for time-integrated solutions. A building-block implementation allows, with the same framework, to predict superficial aerodynamic quantities in distinct conditions: steady, unsteady or dynamic; and in highly nonlinear transonic regimes. The developed methodology is validated in relevant test cases, proving the applicability of our framework to different problems. The computational efficiency against traditional high-fidelity solvers is demonstrated, obtaining computational gains of well over 99.9% on new simulations. Furthermore, a comprehensive comparison between a quasi-steady and a recurrent scheme for forced-motion responses is presented.
This record has no associated files available for download.
More information
e-pub ahead of print date: 8 June 2023
Venue - Dates:
AIAA AVIATION 2023 Forum, , San Diego, United States, 2023-06-12 - 2023-06-16
Identifiers
Local EPrints ID: 484076
URI: http://eprints.soton.ac.uk/id/eprint/484076
PURE UUID: dc9118a2-bd3c-4685-965c-b50673f53534
Catalogue record
Date deposited: 09 Nov 2023 18:13
Last modified: 18 Mar 2024 03:59
Export record
Altmetrics
Contributors
Author:
David Massegur Sampietro
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics