This study leads to many conclusions. First of all Fluent is able to
simulate non-industrial problems which require some particular functions,
such as the Boussinesq approximation for density variation.
Even though this study was not successful, the work done could lead
to another simulation. This deserves to be simulated, starting from what
comes out of this study :
the models have to be large..
3D simulation needs to be done on quite powerful computers with a lot of
memory, to prevent swapping.
The boundary conditions must be periodic, and defined from Fluent, not
from Pre-BFC.
An initialisation of the vertical velocity must be done in order to converge
with very few iterations.
Simulations with a free surface will be available during the year 2000
with the next versionof Fluent (If bought by ENSEIHT)