Download files
Abstract
[Formulae and special characters can only be approximated here. Please see the pdf version of the Abstract for an accurate reproduction.]
Since the damping functions employed by most of the low-Reynolds-number models are related to the non-dimensional distance y+[ special character – near-wall non-dimensional distance in y direction], which is based on local wall shear stress, these models become invalid for separated flows, because the wall shear stress is zero at the reattachment point. In addition, the pressure-velocity correlation term is neglected in most of these models, although this term is shown in this thesis to be important in the near-wall region for simple flows and large pressure gradient flows. In this thesis, two main efforts are made to improve the k – [special character - dissipation rate] model. First, based on Myong and Kasagi’s (1990) low-Reynolds-number model (hereafter referred to as MK model), a more general damping function [special character - turbulent viscosity damping function in LRN turbulent model] is postulated which only depends on the Reynolds numbers [formula – near-wall turbulence Reynolds number]. Second, a form for the pressure-velocity correlation term is postulated based on the Poisson equation for pressure fluctuations.
This modified model predicts the turbulent flow over a flat plate very well. It is found that the inclusion of the pressure-velocity correlation term leads to significant improvement of the prediction of near-wall turbulence kinetic energy. When the model is applied to turbulent flow over a backward-facing step, it produces better predictions than the traditional k – [special character - dissipation rate] model, FLUENT’s two-layer model and the MK model. Again, the pressure-velocity correlation term improves the turbulence kinetic energy prediction in the separated region over that of other models investigated here.
The studies of numerical methods concerning computational domain size and grid spacing reveal that a very large domain size is required for accurate flat plate flow computation. They also show that a fine grid distribution in the near-wall region upstream of the step is necessary for acceptable flow prediction accuracy in the downstream separated region.