EXCELLERATP2 D2.9 Updated Report on the m-AIA Application Use Case

Zenodo (CERN European Organization for Nuclear Research)

Abstract

This document summarises the progress made since Month (M) 12 in the m-AIA application use case (UC) UC-3 within reporting period 2 covering the second year and half of the third year of the EXCELLERAT P2 project.RWTH has advanced large-scale coupled Computational Fluid Dynamics (CFD) – Computational Aeroacoustic (CAA) simulations using the m-AIA framework, starting with a chevron nozzle simulation containing around 1 billion CFD cells and progressing to higher resolutions up to 3.7 billion cells to predict the jet flow and acoustic far-field noise spectra at Mach 0.9 and Reynolds number of 1 million. The corresponding CAA simulation involved up to 4.9 billion degrees of freedom. Far-field noise predictions were enhanced by improvements of the Ffowcs Williams-Hawkings (FWH) solver and it is found that the results are insensitive to the chosen parameters of the FWH surface. The largest simulation was executed on the Hawk High-Performance Computing (HPC) system using up to 4096 nodes achieving high efficiency, showing linear scalability compared to a baseline on 256 nodes. The results show that a full HPC system can be used effectively for aeroacoustic predictions with m-AIA, enabling the execution of large-scale use cases on a pre-exascale level. Workflow automation for m-AIA simulation pipelines was achieved through the development of a Python tool. Visualisation improvements were made using ParaView for handling large datasets and porting activities focused on adapting m-AIA for Graphics Processing Unit (GPU)/Accelerated Processing Unit (APU) architectures. An Artificial Intelligence (AI)-based optimisation workflow is in development in collaboration with project partner Fraunhofer SCAI (FhG). The modular approach supports parallel evaluation of candidate solutions, facilitating efficient optimisation processes for tasks like nozzle shape optimisation.In summary, RWTH has made significant advancements regarding the various aspects relevant to the execution and objectives of UC-3.

Authors 3

1. Ansgar Niemöller Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

2. Matthias H. Meinke Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

3. Arno Feiden

   Fraunhofer Institute for Algorithms and Scientific Computing

   Affiliation as printed

   Fraunhofer Institute for Algorithms and Scientific Computing