Abstract
This paper address the problem of coupling an Adaptive Mesh Refinement method with large-eddy simulation (LES). A multigrid algorithm is used in order to refine the grid in zones detected by a specific sensor based on a wavelet decomposition. This sensor has been developed to accurately detect cells in which the classical LES hypotheses should not be verified. This algorithm is then applied to the simulation of a 3D turbulent mixing layer at Reynolds number 1010. First results exhibit a great sensitivity to the boundary treatments of the interface between the different grid levels. It is shown that the use of a dynamic enrichment procedure allows us to recover results in a very good agreement with non-adaptive very fine grid simulations.