Detached-eddy simulation (DES) is well understood in thin boundary layers, with the turbulence model in its Reynolds-averaged Navier–Stokes (RANS) mode and flattened grid cells, and in regions of massive separation, with the turbulence model in its large-eddy simulation (LES) mode and grid cells close to isotropic. However its initial formulation, denoted DES97 from here on, can exhibit an incorrect behavior in thick boundary layers and shallow separation regions. This behavior begins when the grid spacing parallel to the wall Δ∥ becomes less than the boundary-layer thickness δ, either through grid refinement or boundary-layer thickening. The grid spacing is then fine enough for the DES length scale to follow the LES branch (and therefore lower the eddy viscosity below the RANS level), but resolved Reynolds stresses deriving from velocity fluctuations (“LES content”) have not replaced the modeled Reynolds stresses. LES content may be lacking because the resolution is not fine enough to fully support it, and/or because of delays in its generation by instabilities. The depleted stresses reduce the skin friction, which can lead to premature separation.

/pdf/a-new-version-of-detached-eddy-simulation-resistant-to-yuj0rxxfha.pdf

A new version of detached-eddy simulation, resistant to ambiguous grid densities

P. L. RoeCollege of Aeronautics, Cranfield Institute of Technology,Cranfield MK43 0AL, EnglandIntroductionComputer simulations of fluid flow provide today the sort of detailedinformation concerning special cases that could previously only beobtained from experime.nts. The computer is attractive as a replacement forexperiments that are difficult, dangerous, or expensive, and as an alternativeto experiments that are impossible. Nevertheless, a computer simulationdoes not have quite the same status as a physical experiment because atpresent there usually remains some doubt about its accuracy. Even thoughthe computer code may be free of error to the extent that it operates exactlyas its author intended, it is seldom possible to give a rigorous proof thatthese intentions were in all respects correct. Most of the practical codeswritten to solve complicated problems contain empirical features, some-times in the form of "adjustable constants" whose values must be "tuned"by appeal to the experiments that the simulations are intended to displace.A computer code is described as being "robust" if it has the virtue ofgiving reliable answers to a wide range of problems without needing to beretuned. The ideal code would be one that fully met some declaredspecification of accuracy and problem range, and whose every line was anecessary contribution to that aim. Few codes yet approach that ideal; amajor impediment is that we presently have little idea what properties canbe specified without contradiction.In recent years, however, our understanding of computer codes for aparticular class of problems has advanced some way toward completeness.The problems are sufficiently complex that naive numerical techniques canproduce disaster, yet sufficiently simple that well-understood physics can3370066-4189/86/0115-0337502.00www.annualreviews.org/aronline Annual Reviews

Characteristic-based schemes for the euler equations

A hybrid RANS-LES approach with delayed-DES and wall-modelled LES capabilities

Fluid dynamic turbulence is one of the most challenging computational physics problems because of the extremely wide range of time and space scales involved, the strong nonlinearity of the governing equations, and the many practical and important applications. While most linear fluid instabilities are well understood, the nonlinear interactions among them makes even the relatively simple limit of homogeneous isotropic turbulence difficult to treat physically, mathematically, and computationally. Turbulence is modeled computationally by a two-stage bootstrap process. The first stage, direct numerical simulation, attempts to resolve the relevant physical time and space scales but its application is limited to diffusive flows with a relatively small Reynolds number (Re). Using direct numerical simulation to provide a database, in turn, allows calibration of phenomenological turbulence models for engineering applications. Large eddy simulation incorporates a form of turbulence modeling applicable when the large-scale flows of interest are intrinsically time dependent, thus throwing common statistical models into question. A promising approach to large eddy simulation involves the use of high-resolution monotone computational fluid dynamics algorithms such as flux-corrected transport or the piecewise parabolic method which have intrinsic subgrid turbulence models coupled naturally to the resolved scales in the computed flow. The physical considerations underlying and evidence supporting this monotone integrated large eddy simulation approach are discussed.

New insights into large eddy simulation

Hybrid LES/RANS methods for the simulation of turbulent flows

Progress toward a general purpose hybrid Reynolds-averaged Navier-Stokes (RANS)/large-eddy simulation (LETS) framework is described, in which large-scale, statistically represented turbulence kinetic energy is converted automatically into resolved-scale velocity fluctuations wherever the local mesh resolution is sufficient to support them. Existing hybrid RANS/LES approaches alter the nature of the local partial differential equations according to the local mesh resolution, but they do not alter the nature of the data on which these equations operate. The implications of this are discussed. Subsequently, a simple mechanism is introduced to transfer statistical kinetic energy into resolved-scale fluctuations in a manner that preserves a given set of space/time correlations and set of second moments. This process, which can appropriately be termed Large-Eddy STimulation (LEST), generates the large-scale eddies needed to form the unsteady boundary conditions at RANS interfaces to LES regions, into which turbulence energy can be deposited either through mean convection or through turbulent transport

Interfacing Statistical Turbulence Closures with Large-Eddy Simulation

We evaluate a wall-distance-free low-Re κ-e turbulence closure model which Incorporates an extra source term in the e transport equation designed to increase the level of e in nonequilibrium flow regions, thus reducing the kinetic energy and length scale magnitudes to improve prediction of adverse pressure gradient flows, including those involving separated flows regions. Two such cases are used here to test the model: one in the low speed flow regime, the other a supersonic one. Comparisons with experimental data and with an earlier version of the κ-e model, as well as with a variant of the κ-ω model (both also wall-distance-free) reveal that the proposed model enables improved prediction of adverse pressure gradient flows relative to more traditional κ-e models

A Wall-Distance-Free k-ε Model With Enhanced Near-Wall Treatment

This paper considers the present and possible future state of hybrid turbulence models, which blend statistical Reynolds-Averaged Navier-Stokes (RANS) closures with Large-Eddy Simulation (LES) methods. Experience gathered from existing hybrid formulations is beginning to confirm their ability to bridge the near-wall region in unsteady simulations of high-Reynolds number, separated flow. However, the issue of kinetic energy transfer between regions of differing spatial resolution still limits the range of applicability of the current generation of hybrid methods. A potential for greater generality exists, whereby local grid refinement would automatically lead to a detailed treatment of selected flow regions at a cost lower than that of either conventional LES, or LES with near-wall RANS. Constructing a general hybrid method that interfaces seamlessly between RANS and LES (and vice versa) requires additional considerations relating to spatial/temporal resolution and the transfer of modeled turbulence kinetic energy into resolvable structures. This paper reports some progress in this area. NOMENCLATURE Turbulence kinetic energy, m/s Length scale, m Turbulence production, m/s Reynolds number Strain tensor, Sy Strain magnitude = (2SijSij)Velocity scale, m/s LNS latency parameter Turbulence dissipation rate, m/s Dynamic viscosity coeff., kg/(m.s) Kinematic viscosity coeff., (m/s) Turbulence eddy viscosity , kg/(m.s) Vorticity tensor, Qj . Vorticity magnitude = (2£1^ Qq)' Density, (kg/m) Favre-average of fy Time-average of <j) INTRODUCTION Due to the continued high cost of direct numerical simulation (DNS) and traditional large eddy simulation (LES), engineering predictions of turbulent flow continue to be dominated by simple (single-point, isotropic) ReynoldsAveraged Navier-Stokes (RANS) models. Despite acknowledged uncertainties over the practice of unsteady RANS, these models are frequently entrusted with the task of predicting k L Pk Re S S V

LNS - An approach towards embedded LES

A 'grid-transparent' methodology for CFD

Improved convergence rates for a unified grid framework are achieved by combining several convergence acceleration strategies, which include local implicit time-stepping, low-speed preconditioning, and relaxation methods. It is demonstrated that good convergence can be achieved on various grid types and topologies, all speed regimes, and for both inviscid and viscous flows.

Uriel C. Goldberg

Papers

Interfacing Statistical Turbulence Closures with Large-Eddy Simulation

A Wall-Distance-Free k-ε Model With Enhanced Near-Wall Treatment

LNS - An approach towards embedded LES

A 'grid-transparent' methodology for CFD

Convergence acceleration for unified-grid formulation using preconditioned implicit relaxation