Turbulence models for complex flows.

Industrial devices designed to handle fluid quite often create complex
turbulent flows. This is the case, for example, with turbine engines,
combustors and burners. Unfortunately these flows are still not well
understood and therefore difficult to model. To develop new theories as
well as new models, empirical data are needed. We provide here LDV data for
the cold flow of a double annular burner, for which no database in cold or
hot conditions has been published till now. We propose these data as a new
test case for the validation of turbulence models in complex flows.
The burner was designed to provide a flow with a very good axisymmetry
(less than 2%). Measurements have been taken on a region near the exit of
the burner, on a very refined grid (more than 5.500 grid points). Mean
quantities have been recorded at each location, together with turbulence
intensities and stress. Streamlines of the flow and maps of these
quantities have been provided.
Furthermore, a new and powerful post-processing technique has been proposed
and tested. First, gradients of the velocity vector have been estimated,
giving access to the strain and vorticity tensors. Then the constitutive
equation, which is at the heart of eddy-viscosity turbulence models, and
which links the stress tensor to the latter tensors, has been
experimentally tested. This was done through the computation of invariants
of the flow. Boussinesq's hypothesis, corresponding to a linear
constitutive equation, has been experimentally tested. Several terms
corresponding to a nonlinear development have also been assessed. Finally,
a Prandtl mixing-length and a dissipation rate have been estimated for the
first time for a complex flow. This higly innovative study helps to better
exploit experimental data in order to validate turbulence models.