AI-supported turbulence modelling and simulation for aviation
ROSAS
Credit:
Itsaree - stock.adobe.com
To make aviation more environmentally friendly and competitive, entirely new aircraft concepts and tools are needed that allow for fast and reliable evaluation. In the EU project ROSAS, modern methods of Artificial Intelligence (AI) and Machine Learning (ML) are being combined with the latest developments in Computational Fluid Dynamics (CFD) and coupled with newly developed turbulence models. The aim: to reduce costly and time-consuming tests on real aircraft and significantly accelerate the entire development process.
Within ROSAS, the project partners are developing a new methodology to specifically and realistically investigate the key challenges in fluid dynamics. Using precise computer simulations and targeted experiments, typical flow situations in aviation are replicated under controlled conditions. The resulting data is fed into a comprehensive database, which will primarily support the development and testing of innovative, data-driven AI and ML methods. New approaches for generating computational meshes – i.e., digital representations of aircraft geometry for simulations – will also be tested on this basis. A particular focus is on the further development of so-called turbulence models, which allow for more accurate predictions of complex airflow behaviour. To this end, the project partners are enhancing classical models with AI-based extensions and new theoretical approaches. The goal is to combine (hybridise) different models to significantly increase predictive capability.
Schematic of the FI/ML classic process for data-driven turbulence modeling
In addition, the project defines specific ‘Application Challenge’ test cases – practical use cases closely aligned with configurations of industrial interest – in order to demonstrate and evaluate the new methodologies developed in the project. These cases are closely aligned with the current requirements of the EU's Clean Aviation initiative.
Mach number distribution and stream lines: comparison of the simulation results from the original and
The DLR Institute of Aerodynamics and Flow Technology is focusing in ROSAS on data-driven improvements of turbulence models using AI and ML. An existing model is being adapted so that it can capture physical effects it has not been able to reliably describe before. This adaptation is based on high-quality reference data and involves targeted calibration of complex mathematical terms. In ROSAS, researchers will examine two key flow phenomena that occur around aircraft configurations. For both phenomena, the AI-supported model will be further developed and evaluated so that it can serve as a reliable tool for flow simulation in the future.
ROSAS - RObust simulation Systems exploiting AI based turbulence models and high fidelity algorithmS
Term
1/2025 - 6/2028
Partners
ONERA (Coordinator)
DLR Institute of Aerodynamics and Flow Technology
Università Degli Studi Di Bergamo
Dassault Aviation
Erdyn Consultants
University of Bristol
Barcelona Supercomputing Center
CENAERO
CERFACS
Kungliga Tekniska Höoegskolan
Rolls Royce Plc
Safran SA
Université Catholique De Louvain
Università Di Pisa
Universidad Politecnica De Madrid
Funding
ROSAS project has received funding from the European Climate, Infrastructure and Environment Executive Agency (CINEA) under the Horizon Europe programme under grant agreement no. 101138319.
