Re: [FLASH-USERS] FW: 3D RM Problem

Shuang -

First of all, I must confess and admit that FLASH2.3 will not be ready
so soon as I offered. The most likely dates for the release is late
May/early June. Asking you to wait for so long is, of course, not an
option.

On Tue, Apr 29, 2003 at 06:18:46PM -0400, Shuang Zhang wrote:
> The strange thing to me is, the exactly same setup works fine in
> 2D and 3D 4 AMR levels (one level less). Does this indicate anything?
> Does refine/derefine criterier playing a role?

Usually higher resolution allows for more structure in your flow and
generally more violent evolution. I suspect that the problem is around
boundaries of your grids so modifying refinement criteria (changing
block distribution in your computational volume) will certainly affect
code behaviour.

Using AMR discretization poses certain risk - if you refine too little
some discontinuities will not be captured correctly and stress some
code components.

You may consider rerunning your highest resolution model with less
tolerant refinement criteria. (Adjusting refine_cutoff parameter is
probably the first step.) This will make your simulation less
AMR-like, the runtime is likely to increase, but the model will be
less prone to errors arising from inadequate resolution of the flow
field.

> Also, what is the highest resolution among the successful 3D simulations
> in the FLASH community so far, in your database?

I did not participate in those simulations but some of them had
effective resolution larger than the resolution in your model.

Tomek

--