In conclusion, we could say that we have prefered to test the numerical aspects (meshing, discretisation, multigrid) of the problem rather than the physical phenomenom itself. That is why we did not insist on this in the report.
But the trials we made were not very concluding on this example. In general, results were in the waited direction but sometimes we have had problems to analyse them. And because some of these points are linked with the way Fluent solves the equations, we don't have a lot of information. Moreover, concerning the multigrid option, even if it was quite well detailed in the Fluent's help, it was difficult for us to know which parameters to test.
However it was interesting to have the opportunity to work on this aspects and to get a more practical idea of the importance of the points seen in theory.