Hi Bill,
tried to reproduce what I did yesterday - did not succeed! Can't for my
life remember what I did differently then - today shock seems properly
refined all the way to the boundary.
Must have done something (stupid) yesterday.
Sorry for the noise,
/jon
On 10/14/2016 01:41 PM, Bill Henshaw wrote:
Hi Jon,
Thanks for pointing out a potential problem near walls. I will have to look at it.
I don't recall what "use front tracking 1" does :)
...Bill
On Fri, Oct 14, 2016 at 4:07 AM, Jon Tegner <tegner@xxxxxx <mailto:tegner@xxxxxx>> wrote:
Hi again,
"use front tracking 1" seems to remove this issue! Hopefully this
is true even for the bigger 3D case where I first noticed this.
Regards,
/jon
On 10/13/2016 09:23 PM, Jon Tegner wrote:
Hi,
doing some test with "blast.cmd" (2D), i.e., cgcns with AMR.
Grid is simplest possible, just "rectangle" (and no internal
structures).
Initial data given by "bubbles", just one, but it is placed so
that only a quarter of it is inside the domain (i.e., origin
of bubble sits at one of the corners of the "rectangle").
All four boundaries on mesh are slipWall. This set up should
(I think) generate a solution of a cylindrical shock
propagating through the domain. The issue I see is that when
AMR is on the refinement of the grid is working differently
close to the boundaries, e.g., using nrl=2 the refinement is
lacking over the part of the shock that touches the slipWalls.
This leads to a more "smeared" shock with slightly lower peak
value close to the wall.
When running without AMR, but with a resolution corresponding
to the refined AMR-mesh I don't see the smearing close to the
walls.
Are there possibly some settings regarding AMR one could use
to improve this situation?
Thanks,
/jon