The goal of the present work is a detailed and comprehensive study to assess the accuracy
of the numerical simulation of the mixing processes in a lobed mixer flow field via a
Reynolds-averaged solution method. To meet this goal, the first objective of the current
work was to establish the suitability of various meshing strategies that would allow the
complex mixer geometries found in current gas-turbine engine designs to be captured,
together with the associated convoluted shear layers. A second objective was targeted
at providing further insight and understanding of the capability of eddy-viscosity-based
turbulence models in capturing the convoluted shear layers.
Simplified mixer configurations selected from the literature were studied under incompressible
isothermal flow conditions. Two solution algorithms were employed to
model the mixer flow fields. The first consisted of a pressure-based structured grid
methodology developed for incompressible flows. A density-based mixed-unstructured
grid algorithm for compressible flows was also used, with extensions to low Mach number
flows made possible through a low Mach number preconditioner. The effects of
turbulence were modelled using ak-e turbulence model. The absence of this model in
the code made available for the unstructured algorithm necessitated its implementation
as a first step in the current work.
The effects of unstructured mesh type on the prediction of flows with internal mixing
layers were first assessed for an incompressible planar mixing layer. This simplified case
was used as a benchmark case to help understand the effects on the convoluted shear
layers arising within the lobed mixer flows.
To quantify the capability of a Reynolds-averaged approach in simulating the turbulent
mixer flow field, two variants of the two equation k-e model were employed.
The first constituted the standard linear high Reynolds number k-e model of Launder
and Spalding . The second model was a quadratic non-linear version developed by
Speziale  for the prediction of secondary flows in non-circular ducts. The relative
merits of these two models was assessed through detailed comparisons with experimental
data taken from the literature. Of particular importance in the mixer flow was the
formation and subsequent evolution of the vorticity field. Consequently, this motivated
a detailed study of the evolving vorticity field.
The investigations thus far were based on a simplified mixer configuration with no
temperature differences between the two streams. Therefore, as a final step, a realistic
scarfed mixer was modelled in an attempt to model the temperature mixing.
The main contribution of the present work is the assessment of a grid-based Reynolds-averaged
solution procedure for the prediction of lobed mixer flows. The study revealed
that capturing the initial mixing region proved to be most difficult. Firstly, unstruc-tured meshes employing non-hexahedral elements were very inefficient at simulating the
mixing layer in the early stages. Secondly, the initial mixing region presented significant
difficulties for the Reynolds-averaged solution method in which neither turbulence
model was capable of correctly reproducing the turbulence field. Despite this, global
parameters such as momentum thickness and streamwise circulation were well captured
in the predictions.
A Doctoral Thesis. Submitted in partial fulfillment of the requirements for the award of Doctor of Philosophy of Loughborough University.