[FLASH-USERS] Numerical Effects in Spherical Coordinates

[Michiel Bustraan]

Jason,


The goal of the simulation is to run the shock over larger distances and times (300 days or so). Having a high resolution throughout the domain, which is much larger than the initial shock area, would be difficult.


I do see the value of running the simulation without fluctuations to see the effects. I was made aware in another response that a modification to Grid files can force refinement in an area without introducing fluctuations. I think this might be a less expensive way to see the impact of the fluctuations I introduced.


Either way, a test run with a smaller domain and a higher resolution is a good idea.


Regarding the boundary conditions, a spherical simulation that includes zero radius requires a reflective boundary, does it not?

I could cut out a small area in the center (xmin = 1.0E+12 or so), in which case I would be able to have an outflow.

I'll try that and see if it has an effect.


Thank you for the advice.


Kind Regards,

Michiel


________________________________
From: Jason Galyardt <jason.galyardt at gmail.com>
Sent: Thursday, March 23, 2017 5:37 PM
To: Michiel Bustraan
Cc: flash-users at flash.uchicago.edu
Subject: Re: [FLASH-USERS] Numerical Effects in Spherical Coordinates

Hi Michiel,

If you suspect that your temperature fluctuations are part of the problem, try removing them. You mentioned that you need high resolution; in order to get full refinement, you can set the minimum refinement level equal to the maximum refinement level:

lrefine_min = 10
lrefine_max = 10

Of course, this effectively turns off the AMR feature, so you may have to add processors in order to get the resolution you need across the entire domain. You might also try reducing your domain dimensions for this test run, in order to reduce the cluster resource requirements.

Also, you mention that the gas might be bouncing off the inner boundary: are your boundary conditions reflective, then? If not, there will be no bouncing; if so, turn the boundary conditions to 'outflow' and see what happens.

Cheers,
Jason

------
Jason Galyardt
University of Georgia


On Wed, Mar 22, 2017 at 9:10 AM, Michiel Bustraan <michiel.bustraan at astro.su.se<mailto:michiel.bustraan at astro.su.se>> wrote:

Dear FLASH Users,


I've been running 2D simulations of shocks in spherical simulations.

The initial conditions involve a high density around the origin.


I had to introduce random temperature fluctuations in the inner region to ensure refinement.

I started with quite a high level of refinement, and the maximum level of refinement decreases over time (from 10 layers to 5).


I've found that ripples appear in the inner portion of the domain, as well as larger numerical artifacts.

I think that these might either be due to:


- The initial temperature fluctuations I added, though I feel these would be too small (spatially) to cause all these issues.

- Could it be that the higher pressure region in the center, combined with the lower radial velocity, cause the gas to bounce off the inner boundary?

This might create ripples and internal shocks.


Either way, I'm not certain how to improve my results from here.

If anyone has had similar problems, I would greatly appreciate some advice.


The figures attached highlight my problem.


Kind Regards,

Michiel Bustraan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20170324/c9559da6/attachment.htm>