Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Mechanical Engineering


Tong, Chenning


The effects of the subgrid-scale (SGS) turbulence on the resolvable-scale statistics and the effects of SGS models on large-eddy simulation (LES) are studied. It is shown that the SGS turbulence evolves the resolvable-scale joint probability density function (JPDF) through the conditional means of the SGS stress, the SGS scalar flux, and their production rate, which must be reproduced by the SGS model for LES to predict correctly the one-point resolvable-scale statistics, a primary goal of LES. This necessary condition is used as the basis for studying SGS physics and for testing SGS models. Theoretical predictions, measurements data obtained in a turbulent jet and in a convective atmospheric surface layer, and large-eddy simulation data of convective atmospheric boundary layers are combined to investigate the effects of filter size, the dependence of the SGS turbulence on the flow dynamics, and SGS models performance using new statistical a priori and a posteriori tests developed in this research.
Analyses of the results show that for the inertial-range filter scales SGS predictions of the mean statistics support the premise of LES at the level of lower-order statistics, but not the higher-order statistics. For the energy-containing filter scales the conditional statistics strongly depend on the resolvable-scale velocity and temperature fluctuations, indicating the strong influence of SGS turbulence on the resolvable-scale statistics. The current SGS models have varying levels of performance in predicting different SGS components. The results suggests that efforts to improve SGS models need to ensure that all the relevant SGS fluxes related to the LES statistics are correct predicted. Given the strong dependence of the conditional statistics on the flow dynamics, it may be necessary to incorporate some aspects of the dynamics to correctly predict these conditional statistics. The statistical a posteriori test results are generally consistent with the a priori test results. Similar model strengths and deficiencies are identified in both types of tests. Therefore, analyses of the conditional statistics can serve as an important guide in studying the SGS physics, identifying model deficiencies, and developing improved SGS models.