Institute of Aerodynamics, RWTH Aachen University (Germany)
Local Project ID:
HPC Platform used:
Hazel Hen (HLRS)
A research group from the Institute of Aerodynamics (AIA) of the RWTH Aachen University utilized the computing power of Hazel Hen for large-scale simulations to analyze the intricate wake flow phenomena of space launchers. The objective of the project is the fundamental understanding of the origin of so called buffet loads acting on the nozzle which can lead to critical structural damage and to develop flow control devices to increase the efficiency and reliability of orbital transportation systems necessary for the steadily increasing demand for communication and navigation satellites.
On 12 December 2002, the maiden flight of the European heavy space launcher ARIANE 5 ECA equipped with the new VULCAIN 2 engine failed due to a deterioration of the main stage nozzle. The analysis of the flight data revealed that one of the aggravating factors that led to the accident was the non-exhaustive definition of the loads to which the new VULCAIN 2 engine is subjected during the flight trajectory into space.
The tail of a classical space launcher, e.g., the ARIANE 5, TITAN 4, H-II to name a few, includes an abrupt junction between the main body and the attached rocket engine causing the boundary layer to separate on the base shoulder. At the early transonic phase of the flight the turbulent shear layer shed from the main body impinges on the nozzle just upstream of its end, which due to high dynamic pressure values leads to significant wall pressure fluctuations. Associated with these pressure fluctuations, unsteady aerodynamic forces with pronounced low-frequency peaks, known as the buffet phenomenon, arise which can lead under unfavorable conditions to a complete loss of the space transportation vehicle, as in the aforementioned accident. The steadily increasing demand for communication and navigation satellites and the growing competition in the aerospace market, requires more efficient and reliable transport facilities into the orbit. Therefore, accurate numerical tools validated by high-fidelity experimental investigations are required to provide detailed insight into the wake flow phenomena, to develop methods of their control, and to ultimately reduce aerodynamic loads on the nozzle structure without penalizing the launcher‘s efficiency. Unfortunately, unlike the rocket inner engine flow, the wake flow of a real launcher cannot be analyzed in full-scale on the ground, leading to increased safety margins and consequently, a reduced launcher efficiency.
Within the project, a large number of various configurations including planar space launchers as well as axisymmetric free flight configurations were analyzed at varying freestream (trans-, supersonic) conditions, to improve the understanding of the complex interaction and superposition of different periodic and stochastic flow phenomena, including the not yet fully understood buffet phenomenon. In addition, the effect of passive flow control devices on the wake of a generic planar model was investigated.
The time-resolved numerical simulations of the flow field around the space launcher configurations are performed using a zonal RANS-LES approach. Therefore, the computational domain is split into two zones. In the zones with an attached flow, the Reynolds-averaged Navier-Stokes (RANS) equations are solved and in the wake region a large-eddy simulation (LES) is performed (see Fig. 1). Using such a hybrid approach combined with structured grids an efficient time-resolved computing of high Reynolds-number wake flows at a fraction of the costs of a pure LES is realized.
The flow solver is optimized for the HLRS HPC system using hybrid parallelization based on MPI and OpenMP. Furthermore, parallel I/O procedure using HDF5 is employed. Since the buffet phenomenon is characterized by low frequency pressure oscillations, the required number of time steps per simulation is extremely high to sufficiently resolve these low frequency modes. For example, the analysis of the wake flow of an Ariane 5-like configuration, using a zonal setup with approximately 500 Mio. grid points, a total number of 14.6 Mio. core hours distributed over approximately 10,000 cores have been used. The statistical data requires about 100 TB of disk space.
Project Team and Scientific Contact:
M.Sc. Simon Loosen, Dr.-Ing. Matthias Meinke, Prof. Dr.-Ing. Wolfgang Schröder (PI), Dr.-Ing. Vladimir Statnikov
Institute of Aerodynamics, RWTH Aachen University
Wüllnerstraße 5a, D-52062 Aachen (Germany)
e-mail: office [@] aia.rwth-aachen.de