Keywords: Automotive, Aerospace, Noise prediction, Simulation Workflow, Engineering, Advanced Exascale

Industry sectors: Automotive, Mechanical engineering

Key code used: m-AIA

Organisation involved: Institute of Aerodynamics, RWTH Aachen University

Technical and scientific challenge

Large-scale multiphysics simulations can deliver highly accurate results, but their efficient execution becomes increasingly challenging when a large fraction of a modern HPC system is used. In this use case, a coupled computational fluid dynamics (CFD) and computational aeroacoustics (CAA) simulation was used to predict the sound field of a high-Reynolds-number turbulent air jet emanating from a chevron nozzle.

The simulation involved a very large-scale setup with 3.7 billion cells in the CFD region and 4.9 billion degrees of freedom in the CAA region. When scaled to several thousand nodes on the Hawk HPC system at HLRS, performance bottlenecks and load imbalance became visible. These challenges required specialised expertise in parallel performance engineering and access to pre-exascale HPC resources.

Solution

Runtime performance analysis helped identify bottlenecks in the inter-process communication between the interleaved CFD and CAA solvers. These bottlenecks were not visible in smaller test cases. In addition, the analysis revealed a significant workload imbalance between parallel processes.

To address this, the EXCELLERAT P2 team implemented a dynamic load-balancing approach and optimised communication patterns. This improved the efficiency of the coupled CFD–CAA workflow and enabled better use of the full Hawk HPC system at HLRS.

Impact and benefits

The code enhancements implemented in EXCELLERAT P2 substantially increased the effectiveness of m-AIA for aeroacoustic noise prediction. The full Hawk system with 4096 compute nodes could be used more efficiently, reducing load imbalance and improving runtime performance.

The developments reduced the time-to-solution from days or weeks to less than one day for highly resolved aeroacoustic predictions. This makes demanding engineering studies more feasible and supports faster design iterations.

For a representative large-scale case, the full simulation pipeline can now be completed in under a day. Compared to a baseline on 256 compute nodes, the coupled simulation achieved near-linear scaling, reaching a speedup of 15.9 on 4096 nodes.

This capability supports the industrial development of quieter aircraft and other noise-critical engineering products. It also provides an important foundation for future constrained shape optimisation of chevron nozzles for noise reduction.

Potential EXCELLERAT services

EXCELLERAT can provide expertise in the efficient coupling of CFD and CAA solvers requiring extensive data exchange on HPC systems.

The m-AIA code can be applied to noise prediction and mitigation assessment for engineering problems on HPC systems.

The developed workflow enables highly accurate acoustic field prediction and can support future optimisation frameworks for technical devices.