RObust simulation Systems exploiting AI based turbulence models and hight-fidelity algorithmS

Project Information

ROSAS

Grant agreement ID: 101138319

DOI

10.3030/101138319

EC signature date 2 December 2024

Start date 1 January 2025

End date 30 June 2028

Funded under

• Climate, Energy and Mobility

Total cost

€ 4 630 603,46

EU contribution

€ 4 630 603,46

Coordinated by

OFFICE NATIONAL D'ETUDES ET DE RECHERCHES AEROSPATIALES
France

Periodic Reporting for period 1 - ROSAS (RObust simulation Systems exploiting AI based turbulence models and hight-fidelity algorithmS)

Reporting period: 2025-01-01 to 2026-06-30

Summary of the context and overall objectives of the project

To meet the EU's climate-neutral aviation targets set out in the Clean Aviation Strategic Research and Innovation Agenda (SRIA), a radical transformation of aircraft design is required, introducing new configurations, propulsion systems, and sustainable fuels by 2050. However, this transformation is currently hindered by fragmented digital capabilities across Europe and the limitations of existing Computational Fluid Dynamics (CFD) tools in accurately predicting complex aerodynamic phenomena.
In this context, ROSAS addresses key challenges identified in both the EU Flightpath 2050 strategy and NASA’s CFD Vision 2030. Current CFD methods are limited by high cost, complexity, and insufficient precision for new aviation technologies like hydrogen engines and ultra-efficient wings. ROSAS will hybridize advanced turbulence models, high-order numerical methods, and AI to create next-gen solvers capable of handling complex flow phenomena (e.g. massive flow separation, shock interactions) across the full flight envelope. By doing so, the project will help bridge the gap between academic advances and industrial needs, supporting the flawless entry into service of next-generation sustainable aircraft and reinforcing Europe's leadership in clean aviation.
In the long term, these tools will support the design of climate-neutral aircraft while enhancing EU industrial competitiveness and reducing development costs.
The project activities range from the specification of scientific and application needs, to the development of new tools and models and finally to their validation demonstrating the progress made, and is based more specifically on five main pillars:
• Industrial needs & test case selection – define critical flow problems for training and validation.
• HiFi simulations & experiments – generate high-resolution reference data.
• Numerical methods + AI – merge AI with high-order methods to enhance accuracy, robustness, and efficiency.
• AI-augmented turbulence models – create physics-informed, data-driven RANS/LES models for industrial use.
• Proof-of-concept –demonstrate new CFD tools on real-world aerospace cases, targeting Clean Aviation goals.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

During the first reporting period, significant efforts were dedicated to establishing an efficient project management and coordination framework. Regular meetings and communication tools were implemented to ensure effective collaboration among partners, monitor project progress, manage risks, and facilitate communication with the European Commission and the Consortium.

Substantial progress was achieved across the technical work packages. Industrial configurations and application challenges were defined, while relevant test and training cases were selected and high-fidelity datasets were generated and made available to support the subsequent technical developments. Advanced scale-resolved turbulent flow simulations and experimental activities were carried out, providing the basis for the development and validation of new modelling approaches.

In parallel, significant progress was made in the development and implementation of AI/ML-based methods for turbulence and flow modelling, including the augmentation of existing turbulence models, the development of sub-grid scale and wall models for large-eddy simulations, and the exploration of machine-learning and reinforcement-learning approaches. Strategies for addressing the different application challenges and demonstrating the targeted technology developments were also established, with initial activities already underway.

During the first phase, scientific dissemination and results valorization activities resulted in four peer-reviewed journal publications and ten presentations at international conferences, with further publications and presentations already scheduled.

Finally, communication, exploitation, and data management activities were successfully launched during this initial phase. The project has gained visibility through its website, social media presence, scientific conferences, publications, and other dissemination initiatives; these efforts will be bolstered by an upcoming open ROSAS symposium, which is currently being organized. Exploitation opportunities and key exploitable results have also been identified, while data management and knowledge protection procedures have been established in line with open science principles. Overall, the project has progressed satisfactorily during the reporting period, with no major issues hindering the achievement of its objectives.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

ROSAS will deliver the third generation of CFD tools by integrating AI/ML and high-fidelity simulations to significantly improve aerodynamic and aeroacoustic predictions, reduce computational costs, and accelerate the virtual aircraft and engine design process. Several Key Software Capacities (KSC) will be investigated with ML and developed:
KSC1 : 1HiFi Database of complex and advanced flow simulations
KSC2: New generation of HiFi and AI-based algorithms for accurate and robust CFD
KSC3: AI-based Surrogate models and robust numerical schemes for optimization
KSC4: Coupling methodologies for multiphysics and multi-sources models in HPC
KSC5: AI-based advanced RANS models
KSC6: AI-based advanced WM-LES and LES modelling
KSC7: Advanced workflow in digital simulation process coupled with AI tools
These advances directly support the digital and green transition of the European aviation sector thanks to the following objectives:
Add new datasets to an open database (KSC1)
Aims to improve high-fidelity solvers by ~20%, reducing both computational cost and HPC energy use. Integrate large CFD codes (CODA, elsA/SoNICS, RR Hydra) into design platforms, achieving ≥1% efficiency gains for UHBP fans and centrifugal compressors (KSC2)
Create new surrogate models and ML/HPC tools to boost engineer productivity by ~30% and enable near-real-time prediction across the flight envelope (KS3 & KS4).
Develop advance turbulence modelling for RANS, WM-LES, and LES through selected training test cases, integrating uncertainty quantification into CFD. Achieve ≥90% coverage of the flow conditions envelope, enabling more reliable simulations and more confident design of next-generation aero-engine components (KS5 & KS6).
Improve turbulence modelling (RANS, WM-LES, LES) with uncertainty quantification to correctly predict ~90% of critical flow conditions and increase accuracy on separation-sensitive metrics by ~10% (KSC7).
Consequently, ROSAS’ innovative AI-based methods to create a virtually flying aircraft will allow multidisciplinary designs (such as aerodynamics, structures, and acoustics), using different levels of fidelity, engineers can quickly identify the most promising design solutions with low cost and risk for the manufacturer. These methods will reduce the need for physical prototyping and testing, increasing productivity and allowing more design variants by decreasing time-to-solution in industrial design.