[aspect-devel] Temperature mesh refinement criteria

[John Naliboff]

Hi Felipe,

The mesh-refinement is based on the temperature gradients (more 
specifically, the second derivatives).

 From the original ASPECT paper (Kronbichler et al. 2012, GJI) is the 
following statement regarding temperature-based AMR (see section 3.2.1):

"Finally, we dynamically adapt the mesh every few time steps and we need 
a criterion to decide which cells to refine or coarsen. Since the 
variable that is most indicative of abruptly changing features of the 
solution is the temperature, we apply a criterion to the temperature 
that is commonly referred to as the ‘Kelly error indicator’ (Gago et al.
1983) and that computes for each cell an approximation of the size of 
the second derivatives times the diameter of the cell."

Perhaps some details of the exact approach of have changed since then, 
but I believe this is still the general approach for temperature and 
most of the other refinement criterion.

Cheers,
John

*************************************************
John Naliboff
Assistant Project Scientist, CIG
Earth & Planetary Sciences Dept., UC Davis

On 07/07/2017 12:08 AM, FELIPE ORELLANA ROVIROSA wrote:
>
>  Hello,
>
>         I am doing 3D convection simulations, and finally I have 
> implemented temperature- and strain-rate- based mesh refinement 
> criteria. I tried each one of them and also the merge (hybrid) 
> combination of them.
>
>         After some tests, for my particular problem, I have found 
> better results with the sole temperature-based mesh refinement 
> criterion. But, using this sole T-criterion in lieu of an hybrid 
> T-StrRt, suggests optimal solution of the energy (temperature) 
> equation, and not of the Momentum equation (which would be taken care 
> by the StrainRate criterion). I now both equations are coupled, and 
> perhaps more so in my case where my viscosity is temperature 
> dependent, but I have my queries:
>
>         Is the temperature-based mesh refinement criterion refining 
> the mesh solely according to the temperature itself?
>    or,
>         to the temperature and the temperature gradients?
>
> hope you guys can help me,
> cheers,
> Felipe
>
>
>
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20170707/b6e353d0/attachment.html>