CFC2023

hp-adaptive hybrid RANS/LES simulations using a high-order discontinuous Galerkin method on unstructured meshes

  • Basile, Francesca (ONERA)
  • Chapelier, Jean-Baptiste (ONERA)
  • Laraufie, Romain (Airbus Operations SAS)
  • Frey, Pascal (Sorbonne Universit√©)

Please login to view abstract download link

In this work, an hp-adaptation approach for unstructured meshes based on high-order discontinuous Galerkin (DG) schemes is applied to hybrid RANS/LES simulations. The strategy for h-adaptation of tetrahedral elements consists in performing metric-based remeshing using MMG3D [Dapogny2014], while prismatic elements can be only enriched in p. The adapted h and p maps are prescribed using an a posteriori error estimator combining the measure of the energy of the highest order modes and the inter-elements jumps of the DG solution. The choice of refining either in h or p is made by evaluating a smoothness indicator based on the decay rate of the DG modal coefficients. The DG flow solver employed in this study is the software CODA [Leicht2016], the new generation CFD platform developed jointly by ONERA, DLR and Airbus. The hp-adaptive algorithm is first tested on the subcritical flow past a sphere, at Reynolds number 3700 using Detached Eddy Simulation. The flow solution obtained on the resulting hp-adapted mesh matches LES numerical results from the literature, while employing a lower number of degrees of freedom (dofs). Some discrepancies with the DNS data show the strong dependency of the hp-adaptive process on the employed turbulence model. Then, hp-adapted DG computations of the jet issued from the PPRIME nozzle at Reynolds number 1 million [Bres2018] are carried out with a Zonal Detached Eddy Simulation mode 0+1 approach, and the impact of the increased resolution is analyzed for three adapted meshes presenting up to 32 million dofs [Basile2022]. Qualitative results compare fairly well to numerical references from the literature obtained with FV schemes on structured meshes and non-adaptive DG methods, with a reduced number of dofs. An acoustic analysis of the far-field propagated jet noise is performed on the hp-adapted meshes using the FW-H method implemented in the acoustic flow solver KIM, providing noise levels in good agreement between the adaptive simulations and the numerical and experimental references. However, some discrepancies are found in the flow field obtained on the finest adapted meshes, with respect to the results obtained in the experiment. In particular, longer potential cores, associated with lower turbulence intensities, are found when increasing the resolution. This appears to be a recurrent issue in the literature, especially when dealing with simulations in which the turbulence is not explicitly triggered inside the nozzle.