Principal Investigator: Wolfgang Schröder, Institute of Aerodynamics, RWTH Aachen University
A research team of the Institute of Aerodynamics (AIA) of the RWTH Aachen University leveraged the petascale computing power of the HPC system Hornet for large-scale simulation runs which used the entirety of the system’s available 94,646 compute cores. The project “Large-Eddy Simulation of a Helicopter Engine Jet” aimed at analysing the impact of internal perturbations due to geometric variations on the flow field and the acoustic field of a helicopter engine jet. For this purpose, the researchers conducted highly resolved large-eddy simulations based on hierarchically refined Cartesian meshes up to 1 billion cells over a time span of 300 hours.
Noise reduction is one of the major tasks of today’s aircraft development and is also one of the key goals in European aircraft policy. Compared to 2000, the perceived noise level of a flying aircraft is supposed to be reduced by 65% until 2050. Especially during take-off, jet noise is the main contributor to the overall acoustic field. To comply with new noise level regulations, reliable, efficient, and accurate aeroacoustic predictions are required which are determined by the quality of the resolution of the turbulent flow field. To simulate a jet at realistic high Reynolds number conditions by a large-eddy simulation (LES), it is required to have a mesh resolution on the order of more than O(109) mesh points. It is the intricacy of the flow structure that requires such a detailed resolution which can only be realized on an architecture like, e.g., the Hornet of the High Performance Computing Center Stuttgart (HLRS).
The current research work is conducted within the European project Coupled Parallel Simulation of Gas Turbines (COPA-GT), in which a full turbine of a turbo-shaft engine is simulated. The major project goals are to analyze the flow for a complete non-generic nozzle geometry and to use hardly any modeling for the physical processes so that the results for the acoustic field are not contaminated by modeling errors and can be used to validate and/or calibrate simulation methods that describe turbulence noise sources.
In this large scale computing project, the inside nozzle flow plus the jet flow are computed on the HLRS system Hornet by a monotone implicit LES approach using a high mesh resolution of approx. 1.1 billion cells. The required average machine time is 300 hours, i.e. 3 million core hours. The minimum number of processors for the flow simulation is 10,000. However, for scaling purposes the computations are run on up to 94,646 processors. The statistical data requires approx. 120 TB of disk space. The flow field is qualitatively illustrated by the instantaneous vorticity field in the rear part of the nozzle and the jet region. The mosaic-like structure of the vorticity distribution evidences the intricacy of the flow field and gives an idea of its multi-scale character.