THE EFFECT OF DIFFERENT GRIDS AND DIFFERENT LES MODELS ON TURBULENCE STATISTICS IN CHANNEL FLOWS

摘要

The present paper looks in detail at how different types of relatively coarse numerical grids relevant to industrial applications and different types of large eddy simulation (LES) models influence turbulence statistics in the near wall region. The behaviour of both the grids and the LES models in this region affects the wall shear stress and near wall turbulence both of which are of critical importance for studies of heat transfer. Channel flow is used as the reference flow field since there is now a large amount of DNS data available for comparison. A number of different types of grids are studied including uniform, stretched and hybrid grids. The LES models studied are; the classical Smagorinsky model [Smagorinsky 1963], Dynamic Smagorinsky model, and the WALE model [Nicoud 1999]. Comparison of energy budgets including the LES model subgrid contributions reveals that that the WALE model is highly active even for under resolved flows while both the Smagorinsky and dynamic Smagorinsky models are not really implicated. The present study also shows that unstructured grids based on tetrahedral cells induce large numerical correlated noise that shows even in the averaged variables.