Still today, the numerical representation of a fully 3D turbulent flow remains one of the most challenging task. In Computational Fluid Dynamics (CFD), turbulent flows can be numerically solved with different levels of accuracy bounded between Reynolds Averaged Navier Stokes (RANS) and Direct Numerical Simulation (DNS) methods. Today, the RANS is the standard approach to perform turbulent flow simulations. It allows a good reproduction of the mean flow conditions guaranteeing, at the same time, an acceptable computational cost for practical engineering applications. Unlike the RANS, DNS is the most complete approach that can be used to numerically solve a turbulent flow because, in this case, all the turbulent scales are directly solved. However, today the DNS approach remains inapplicable in industrial field because of the prohibitive computational power required. Between RANS and DNS, a third method can be considered for the solution of high Reynolds turbulent flows: Large Eddy Simulation (LES). LES approach allows the direct solution of the largest turbulent scales (anisotropy turbulence) while the smallest scales (isotropy turbulence) are numerically modelled by a sub-grid scale model. For this reason, with respect to RANS, LES is expected to give an improvement about turbulent flow numerical solution when the physical behaviour of the considered fluid domain is dominated by the large scales of motion. At the same time, the computational cost of a LES simulation is quite lower than for DNS simulation. Even if during the last years LES has helped to improve the comprehension of complex turbulent fluid dynamic systems, for its applications in industrial fields further insights are needed. In particular, one of the main problem linked to the LES method regards the definition of a simulation methodology by which to obtain an high solution level (i.e. high level of energy directly solved) at the lowest computational cost. To fulfill this requirement, the authors defined a new LES simulation methodology based on Adaptive Mesh Refinement (AMR). The first results obtained by its application on a backward facing step test case are here presented and discussed in detail.

This content is only available via PDF.
You do not currently have access to this content.