[FLASH-USERS] Pressure Fluctuations from AMR

[Tomasz Plewa]

Hi Michiel -

What I mean by on-and-off refinement is when a fine mesh patch is 
removed in a part of the solution that is deemed sufficiently smooth for 
de-refinement. It is possible that solution in the same, now coarsened 
region may appear non-smooth at the next step, and so the region will 
get refined, it will be found smooth and de-refined, and so on. The 
problem is rarely related to actual physics and more often to numerical 
errors occurring at the fine-coarse boundaries of de-refined regions.

Tightening your refinement criteria may cure the issue, although it may 
not be needed for the whole set of solution variables. Trial and error, 
methinks.

Tomek
--
On 09/13/17 17:15, Michiel Bustraan wrote:
> Tomek,
>
> Could you expand on the difference between what you call on-and-off and uniform refinement?
> Do you mean that this wouldn't happen with a uniform resolution, because that is correct.
> In the absence of refinement, the simulation runs normally.
>
> - Michiel
> ________________________________________
> From: flash-users-bounces at flash.uchicago.edu <flash-users-bounces at flash.uchicago.edu> on behalf of Tomasz Plewa <tplewa at fsu.edu>
> Sent: Wednesday, September 13, 2017 7:12 PM
> To: flash-users at flash.uchicago.edu
> Subject: Re: [FLASH-USERS] Pressure Fluctuations from AMR
>
> I bet on on-and-off refinement/derefinment. I would expect no such
> features appear in uniformly refined models.
>
> Tomek
> --
> On 09/12/17 08:25, Michiel Bustraan wrote:
>> Dear FLASH users,
>>
>>
>> I've been encountering a problem with AMR.
>>
>> The attached images show the issue.
>>
>> I always find sharp pressure fluctuations occuring near the origin,
>> and this only happens with several layers of AMR.
>>
>> Since this region does not contain high velocities or pressures, it
>> seems bizarre that this is happening.
>>
>>
>> These simulations use spherical coordinates, 2D, and split PPM hydro.
>>
>>
>> I originally ran the simulations without specialized grid refinement
>> conditions, which led to worse pressure spikes.
>>
>> I was hoping that by layering the refinement (5 layers out to 2e15 cm,
>> 4 layers to 4e15, etc.) using Grid_markRefineDerefine.F90,
>>
>> it would solve the issue. However, as you can see in the attached
>> images, that has not been the case.
>>
>>
>> Has anyone experienced something like this using AMR?
>>
>> Since the region is fairly flat (in density, pressure, and
>> temperature), I am uncertain where these features could come from.
>>
>>
>> Kind Regards,
>>
>> Michiel Bustraan
>>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: tplewa.vcf
Type: text/x-vcard
Size: 425 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20170913/54de2924/attachment.vcf>