We think that we succed to create a realistic geometry and a unstructured mesh, which was a great part of the work. It's a pity that we have obtained strange results for the simulations even if the global behaviour of the flow in downstream of the bubble and the sphere is in "good" agreements with the theory. We realize that with a industrial code even if study flow seems not too difficult it is not always easy to obtain acceptable results.
Nethertheless, this task allowed us to manipulate prebfc and to improve our level in the using of the industrial code Fluent.

We  suggest futur tasks to resolve these problems:

                    - use of an uncartesian structured mesh
                - 3D study                 - use of the USER DEFINE FUNCTION to define correct boundaries conditions