Universality of large eddy simulation model parameters across a turbulent wake behind a heated cylinder

Abstract

The issue of universality in the specific context of subgrid scale (SGS) physics and modelling for large-eddy simulation (LES) is addressed in this study from two perspectives, i.e. global averages and conditional averages. The specific objectives of the present a priori study are to examine how uniform (i.e. universal) are model parameters across a generic shear flow, both in terms of the mean flow and in terms of the coherent structures. Other related objectives are to document possible differences between momentum and scalar SGS dissipations, and to compare eddy-diffusion and nonlinear model predictions. Measurements are performed across a high Reynolds number heated wake flow, using an array of four X-wire and four cold-wire probes. This array enables us to obtain two-dimensional filtered quantities in the cross-stream and streamwise directions by invoking Taylor's hypothesis with three different filter scales. The results show that there are significant differences between the general trends in globally averaged profiles of SGS kinetic energy and scalar-variance dissipations. The peak levels and peak locations of the two dissipations are quite different. The model coefficients evaluated from matching real and modelled SGS dissipations in the globally averaged dynamics show that the model parameters are quite universal, both for velocity and passive scalar fields, and also for the various models considered. The phase-averaged results (conditioned upon the passage of von Karman vortices) reveal that there is a strong effect of the large-scale structures on SGS dynamics, and that the Smagorinsky model parameter is not universal across different regions of the coherent structures. Generally, the mixed nonlinear model provides better predictions of kinetic energy and scalar-variance dissipations than the Smagorinsky model, i.e. model parameters associated with the mixed model are significantly more universal.