High-Fidelity CFD Simulations of a Full Open-Fan Geometry with an Improved Overset Grid Method
Please login to view abstract download link
This paper introduces a new methodology to perform high-fidelity Computational Fluid Dynamics (CFD) simulations around rotor geometries in open-air configurations. This approach consists of representing the near-body regions with structured curvilinear meshes that are automatically extruded from each geometry surface mesh. These meshes are then immersed in a Cartesian grid consisting of an isotropic core around the geometry and anisotropic grids elsewhere. This topology reduces the number of cells in the Cartesian background grid while ensuring mesh conformity. Inter-grid connectivity is managed using a second-order overset method with static and dynamic data transfers. This methodology has been implemented in Cassiopee (CFD Advanced Set of Services in an Open Python Environment) and was recently improved to perform simulations with hubs or spinners via a parallel double-wall algorithm. The overset method has also been optimized for better performance and shorter computation times per iteration. These developments have been validated for various rotor, propeller, and mini-drone geometries. In this study, the methodology is applied to a generic Open-Fan geometry under acoustic take-off conditions using the configuration and operating points provided by Safran Aircraft Engines (SAE). In addition to Cassiopee functions, the simulations are carried out using the HPC finite-volume structured-grid-based FastS solver coupled with a Zonal Detached Eddy Simulation (ZDES) turbulence modeling strategy for the entire computational domain. High-fidelity aerodynamic results are obtained in a short amount of time, showing good to excellent agreement with URANS reference simulations for averaged quantities. This industrial-level case study demonstrates the adaptability of the current approach and the simplified management of the parallel workflow from pre-processing to post-processing. Finally, both accuracy and performance are assessed on three different meshes with cell counts ranging from one billion to three billion, using finer grid resolutions in the near-body and off-body flow regions. Particular emphasis is placed on capturing turbulent structures in the regions of interest.
