An approach to wall modeling in large-eddy simulations

Abstract

Channel flow with friction Reynolds number Reτ as high as 80 000 is treated by large-eddy simulation at a moderate cost, using the subgrid-scale model designed for detached-eddy simulations. It includes wall modeling, and was not adjusted for this flow. The grid count scales with the logarithm of the Reynolds number. Three independent codes are in fair agreement with each other. Reynolds-number variations and grid refinement cause trades between viscous, modeled, and resolved shear stresses. The skin-friction coefficient is too low, on the order of 15%. The velocity profiles contain a “modeled” logarithmic layer near the wall and some suggest a “resolved” logarithmic layer farther up, but the two layers have a mismatch of several units in U+.

Reynolds Stress Structures in the Hybrid RANS/LES of a Planar Channel

“…Such a constraint forces the closure to act in RANS mode near the wall, through the sensitivity in ∆. Even with a ∆ + x > 8000, Nikitin et al [23] showed that turbulence is sustained in the bulk of a channel flow, lending support for the method. Despite this, hybrid RANS/LES has been shown by numerous authors to be sensitive to the streamwise grid spacing [3,21,23,24,25].…”

A joint numerical and experimental study of the jet of an aircraft engine installation with advanced techniques

“…More precisely, if the switch in LES mode occurs inside the RANS boundary layer, this will result in an underestimation of the skin friction coe©cient [10]. The grey-area needs careful monitoring but novel approaches are emerging [11,12] to overcome this limitation.…”

Large eddy simulations in 2030 and beyond

“…In the presence of strong favourable or adverse pressure gradients, in separated flows or in flows in which the mean velocity is highly three-dimensional, hybrid methods have found application to WMLES as well. Several strategies can be used to switch between RANS in the wall layer and LES in the outer part of the boundary layer, such as changing the length scale in the model from a RANS mixing length to one related to the grid size [39], assigning the RANS and LES zones a priori [40] or using a blending function to merge the SGS and RANS eddy viscosities [41,42]. A more complete review can be found in [36].…”