AbstractA new quadratic k−ω turbulence model with enhanced treatment for near-wall turbulence anisotropy is developed, implemented, and validated in this work. The model abandons the linear Boussinesq approximation and employs an expanded tensor polynomial formulation for the Reynolds stress tensor, or its corresponding anisotropy, applying it to the standard k −ω formulation. The original transport equations for the turbulence scales are retained. The modification for the near-wall turbulence anisotropy uses a novel approach which considers how anisotropic turbulence effects are manifested in different regions of the boundary layer. This is achieved using damping functions that rely only on local variables, specifically turbulence quantities and velocity gradients. The formulation is incorporated into the developed non-linear k − ω framework through its anisotropy expansion coefficients.
Initially, the model predictions on simple boundary layer flows, namely a fully-developed plane channel flow at various Reynolds numbers and a flow over a flat plate, are analysed to validate the formulation and implementation of the model. The performance of the model on a configuration that involves streamline curvatures and internal shear layers, specifically a curved channel flow, is also examined. The model is subsequently applied to more complex configurations that exhibit features such as separation, recirculation, impingement, and swirl: a planar diffuser with a downstream monolith and a swirling flow through a sudden expansion. The model shows to be effective in returning improved predictions, relative to the standard k − ω model, for many aspects of the flows presented. For example, on the prediction of pressure losses on the curved channel and on the prediction of reattachment points on the sudden expansion. The maximum increase in computation time associated with the model is around 40% compared to the linear k − ω model. Nevertheless, the new model is shown to be robust in the configurations tested, i.e. stable and converged solutions are obtained and no change in the numerical or computational setup is needed. Based on the cases considered, the model is expected to be particularly advantageous in applications involving internal flows in which the abovementioned flow features (e.g. separation, recirculation, and impingement) are present, for example, for modelling flows in automotive exhaust catalytic converters, fuel systems, or turbochargers. While room for further development is identified, for example by adding an explicit rotational correction, the general conclusions of this work are encouraging.
|Date of Award||Apr 2020|
|Supervisor||Svetlana Aleksandrova (Supervisor), Humberto Medina (Supervisor) & Stephen Benjamin (Supervisor)|
Student thesis: Doctoral Thesis › Doctor of Philosophy