I use unstructured mesh to deal with my problem because of the complexity geomery. But the mesh quality is not perfect and contains many skewed meshs. Finally it make the solution converge difficultly. How to solve this problem and get ideal solution??

Thanks!

humm, thats a good question, i can only give u little insite to what happens while solving a prob, as i understand, (the main problematic area) when u solve for pressure correction ..it is done in two steps first a pressure correction pdash is solved neglecting nonorhtoganl terms that is assumign that grids are perfectly orthogonal , since here we intorduced incorrection we solve another pressure correctiong to make up for nonorthogonal grids, this is called second pressure corrector step, in theory u can solve another pressure correction step and so on,

usually it is set to one step only, if u feel ur mesh is too skew then u can set it to more steps giving better results, (for more detail see book by peric, Computational methods for fluid dynamics, chapter 8th)

second place u can get improvement is: by default (as i think) fluent uses algebraic multigrid (AMG) which i s good and easy to solve rather than using FAS (full approximation storage) multigrid, here u store everything (rather than approximating) and is beeter than AMG for skewed meshed, try using this will help.

You can adapt the boundaries and the volume so that skewed elements are eliminated. How you adapt the boundaries depends on what you are working on.

I think you can work in two different aspects: First you can improve your grid quality by using the better grid generator or by changing some strategies in grid generation. For example you may divide your geometry to some more regular base areas and produce mesh for each of them separately. In the second aspect, you can solve convergency problem from solver (FLUENT itself). If you have complex geometry and using the unstructured grid generation, to get a accurate results you have to use high order discretization which will lead to difficulty in convergency. In such cases reducing relaxation factors may help to have stable convergency. Also you can solve the problem first by First Order discretization and after convergency increase the order of discretization and again continue to iterate.