Issues and case studies for experimental validation of LES in incompressible
Charles MeneveauMechanical Engineering and Center for Environmental and Applied Fluid Mechanics
The Johns Hopkins University, Baltimore - USA
Monday, June 20, 11:00 CERFACS Conference Room
The Department Numerical Methods, with offices in Braunschweig and Goettingen, is responsible for the development of methods for fluid flow simulation. The department primary task, therefore, is to provide robust and flexible numerical tools for the efficient and precise analysis of proposed Aerodynamic configurations.
This presentation will review several experimental data sets obtained for the specific purpose of detailed testing of LES models and codes. We review a recent remake of the Comte-Bellot & Corrsin decaying grid turbulence experiment at higher Reynolds number [1], where the data consist of filtered velocity statistics including higher-order moments and PDFs. We also discuss measurements of spectral eddy diffusivity obtained in a heated cylinder wake [2] and describe the methodology to use thermal anemometry, multi-probe data to measure the spectral distributions of SGS dissipation. The third case study arises in recent experiments in rapidly strained flows, in which we present several a-priori studies of energy fluxes and attempt to verify empirically whether LES-defined variables tend to RANS-like variables as the LES filter scale becomes large. Measured energy fluxes (subgrid dissipation) at various scales and comparisons to the Reynolds production show that for a homogeneous flow, LES variables indeed tend to their RANS counterparts. These results are obtained based on experimental data from the JHU planar straining facility [3]. Conversely, for a non-homogeneous flow (in the near wake behind turbomachinery blades), the results show that the Reynolds production differs significantly from the mean subgrid dissipation. These results are obtained in the JHU axial turbomachine facility [4]. Specifically, we argue that the paradoxical observations require a new dual decomposition of kinetic energies into 4 terms, and the fluxes among them involve five different quantities [5]. Specifically, the important role played by mean subgrid velocity field is due to the spatial filter size being comparable to the characteristic mean velocity length scales (about a third of the wake width).
Cerfacs' Conferences 2004-2005 Home Page
