Hybrid RANS-LES Methods

Abstract These approaches to the numerical simulation of high-Reynolds-number turbulent flows combine Reynolds-Averaged Navier–Stokes (RANS) elements and Large-Eddy Simulation (LES) elements. This combination can take many forms, and a classification is proposed here along with the various distinctions that need to be made. First, the reasons not to rely on either pure RANS or pure LES, unfortunately leading to complexity and increasing the user of HRLM's burden, are summarized. The primary applications are to transportation. We then compare “RANS-before-LES” and “RANS-under-LES” approaches. We recall that some approaches have the ambition of covering the entire spectrum from Direct Numerical Simulation (DNS) to RANS, but most approaches do not. The length scale in the model can be based on any one of the following: the size of the grid cell only; the cell size and the flow field; or only the model and flow field. Each comes in various versions. Methods can be zonal, with the user dictating the approach in each region, or seamless, with a single equation set making the best use of the grid locally. RANS-under-LES methods suffer from Log-Layer Mismatch, which must also be controlled in all Wall-Modeled LES (WMLES) methods. It is argued that, with current computing power, WMLES methods can be regarded as applying RANS under LES and often RANS before LES, so that, although the claim is often made, they do not significantly by-pass the fundamental empiricism of RANS. The review ends with a brief discussion of practices for grid generation, time integration, and flow solution.