I'm currently performing a 2D grid convergence study using cfx4.3. I'm having trouble generating fine grids in build. When I get down to around y+=1 equivalencing misses out a lot of nodes. Then when I try to manually equivalence I just end up collapsing near-wall elements. I've tried working in a larger scale but without any success. Any tips?

Cheers, Steve

Steve,
Perhaps you should take a look at your global tolerance value in the geometry. If it is too large, build adjusts it automaticly for you, otherwise usually you get errors like those you mentioned.
Regards,
cfd guy

I've played around with the tolerance quite a bit and the adjusted tolerance (by build) is always very very small since my grid is small. As a consequence some nodes are not deleted during equivalencing. I had thought that rounding errors were my problem but when I redraw the block structure in micrometres instead of metres I get the same problem.

Thanks for your help, steve

(1). I had a lot of problems when using commercial codes to generate fine meshes, this included Fluent family, CFX-TASCflow family, ICEMCFD. (2). Since I don't know exactly how these codes create meshes from the geometry, it is hard for me to give advices. (3). My feeling is: the mesh is not created directly from known geometry, and there are some kinds of interpolation used in the process. (4). In principle, if you generate the geometry and the mesh together, you should be able to create mesh with no problem at all. (5). But if you create the geometry separately, and then try to generate a mesh, it is likely that you will have to use some kind of projection or interpolation. In many cases, the grid cross-over will result in mesh failure, if the two mesh line are very close together. (6). This is the place where users should try to write his own code to model the geometry and the mesh at the same time. (easy to say but hard to do I guess) In other words, mapping a mesh onto a geometry will likely create troubles,while it should be all right, if you calculate the mesh directly on the geometry.

When your geometry is simple, define your geometry in USRGRD. Then there is no limit but, yes I know, there is some work to do.

Astrid

I think your problem is that you have some regions of very small cells in a relatively large geometry.

To get around this problem you should equivalence the problem area manually with a large tolerance, EXCLUDING the nodes which will be collapsed (the smallest cells).

This should get around your problem.

I agree, and have been equivalencing manually. Using this method I can equivalence the coarse cells OK. However equivalencing still fails when I try to equivalence the near-wall cells - where of course I need a small tolerance. Perhaps I need to try a different grid generator.

Thanks for your help, Steve