Hello, I would like to ask your help about the following problem I encountered using GAMBIT 2.04 and Fluent 6.1.18. I have a 2D axisymmetric relative complex internal geometry for a flow at very low Reynolds numbers (150-500), and in order to economize memory and CPU time I would like to use a structured mesh. I use Map mesh schemes, dividing my geometry in logical rectangles, and I use non-uniform node spacings for edge meshes. Let's assume I have two horizontal edges of different length, AB and BC: they share the vertex B and I put respectively 20 and 60 intervals on each. Using different values of "Successive ratio" in the panel "Mesh Edge", I make sure that the last interval on AB and the first interval on BC are of the same order of magnitude. I build up all my mesh this way, thus obtaining a very regular mesh and nondissipative mesh: correspondingly, the convergence rate of the simulation in Fluent is very good. I have troubles when refining the mesh for grid indipendence studies. If I double the number of intervals on each edge, i.e. 40 intervals for AB and 120 for BC, then, due to the non-uniform node spacing on AB and BC, the "boundary" intervals won't be of the same order of magnitude anymore. This is detrimental to the quality of the mesh, giving rise to much worst convergence rates than 4 x the original convergence rate, or even making convergence impossible at all. Does anyone here know how to solve this problem? It doesn't arise if I use uniform mesh spacings, but even at my very low Reynolds numbers it would be too much of a limitation. I would also like to hear form other Fluent users about their experience with structured mesh refinement.

Thank you very much for your attention,

Andrea Panizza

i think u should use the unstructured meshing