Fast parallel remeshing for accurate large-eddy simulations on very large meshes

Numerical simulations on very large meshes, such as large-addy simulations (LES), cannot be performed without resorting to distributed-memory parallelism. For these methods, a sufficient precision can only be achieved by remeshing dynamically the areas that need it. Such a remeshing must therefore be performed in parallel. This paper presents the coarse-grain parallel remeshing method which has been devised and implemented in the PaMPA library for handling distributed meshes in parallel. This method is validated in the context of an industrial LES simulation on a helicopter turbine combustion chamber, on a mesh of more than one billion elements.