Periodic Reporting for period 2 - NextSim (CODA: Next generation of industrial aerodynamic simulation code)
Periodo di rendicontazione: 2022-09-01 al 2024-02-29
Aviation contributes to more than 2% of global greenhouse gas (GHG) emissions and its activity is increasing exponentially. In the absence of further measures, carbon dioxide (CO2) emissions from international aviation are estimated to almost quadruple by 2050 compared to 20101. In the process of obtaining new and more efficient products, numerical simulation (such as Computational Fluid Dynamics, CFD) is becoming a key player in aeronautical design. However, in despite of the current deployment CFD during the design process, NextSim partners, recognise that there is a need to increase the capabilities of current numerical simulation tools for aeronautical design by re-engineering them for extreme-scale parallel computing platforms. By doing so, the aerospace industry will be able to expand the use of HPC (High-Performance Computing) in the design loop. NextSim aims at improving industrial applicability of CFD software, in particular CODA, on leading-edge HPC systems. The CODA solver includes classic finite volume capabilities and new highly accurate high order discontinuous Galerkin schemes, all specifically tailored for aeronautical applications, and available to all partners. The CODA solver will be the new reference solver for aerodynamic applications inside AIRBUS group (including aircraft and helicopters), as such, it will be used by AIRBUS for the design of a variety of engineering applications, having a significant impact in the aeronautical market. This targets the efficient implementation on modern HPC hardware as well as extended functionalities making use of the available HPC resources, in particular turbulence scale-resolving simulation capabilities which would not be viable without the extensive computational resources offered by massively parallel HPC cluster environments. NextSim specifically aims at both algorithmic/numerical improvements and HPC aspects of their implementation at the same time, which can be conflicting goals in some cases.
All the activities and tasks that were initiated in the first half of NextSim have enjoyed great improvements and led to successful results, summarized in the following lines. WP1 did the early selection and definition of common test cases that are complex enough to provide reasonable information on actual industrial application performance but at the same time accessible to all beneficiaries interested in considering them has proven essential for the fruitful exchange and comparison of experience gained throughout the project. The definition of a common appropriate test suite has made sure that research activities and software developments have been tested “early enough” on the challenging cases considered here rather than just on idealized simpler cases as often required during the actual development. In addition, the WP1 test case focused Best Practice Guidelines report will continue to serve as a major dissemination means towards the CODA community. After, in WP2 all the fruitful collaborations started at the beginning of the project continued in the second part. BSC took advantage of the DGSEM-related developments provided by ONERA and CERFACS and build upon them to try to identify best practices for high-order LES simulations (reflected in WP1 summary). The multigrid developments are an excellent example of a collaborative effort, too. ONERA and DLR have developed their respective parts of the solver (agglomerator and solution algorithms) and worked jointly during the testing phase for both. In WP3, several partners have investigated and developed several feature detection techniques, such as (Spectral) Proper Orthogonal or (High Order) Dynamic mode decomposition, in a way that will allow them to become regular tools in the analysis of fluid flows. During the following months, those tools would be readily available and will be used in the analysis of the data produced in WP1, providing valuable information of the physics involved in those complex configuration and generating new models for improved visualization and optimization. Furthermore, in WP4 have been progressing as expected, with a significant impact into the development of CODA. This include the development of performance models for the relevant kernels, performance analysis of the baseline CODA version in NextSim, improvements on the relevant kernels in current generation architectures, development of mini apps to assess the capabilities of CODA limiting kernels on novel EuroHPC architectures, development of novel load balancing algorithms, dynamic balancing strategies and the assessment of communication pasterns in CPU-GPU platforms. Finally, in WP5 Dissemination and exploitation activities have been planned to cover the full length of the project, starting from M1, till the end. More particularly, dissemination activities are the one initially carried out, with the development of the website and the creation of the social media accounts, while exploitation and impact assessment are planned by the end of the project. The project consortium is fully aware of the importance of these activities and have been working to fulfil the expectations from the first day. As overall assessment of the work done, one can consider that the project has reached the target, while in the coming months after the closure, the impact of the activities could further spread to other research communities.
We increased the capabilities of current aerodynamical numerical simulation tools for extreme-scale parallel computing platforms. This is particularly important and revolutionary for aircrafts, where more simulation and less testing will have strong impact in the process chain "from design the prototype to mass production”.