[aspect-devel] Hanging nodes problem

[Lev Karatun]

Hi John,

so I looked at the model for which I had the strainrate strategy enabled
alongside with composition. It looks like high strainrate is in the cells
that are already at max refinement level, so they don't get refined. The
error indicator of the neighboring cells is low. I attached the screenshots
to illustrate it.
Another thing I noticed is that there is no anomaly at the last timestep
that converged successfully. Every time I had this problem, the same
pattern occured:

timesteps ... - ***0  -- velocity anomaly
***mesh refinement occurs***
timestep ***1 -- no velocity anomaly, successful convergence
timestep ***2 -- non-convergence
(hope it's not too confusing)

Best regards,
Lev Karatun.

2017-05-19 11:22 GMT-04:00 Lev Karatun <lev.karatun at gmail.com>:

> Hi Anne,
>
> yes, I updated Aspect 4 days ago, and DealII about 2 weeks ago.
>
> John, strainrate-based refinement is what also came to my mind when I
> first faced the problem, and I tried it about a month ago. It also didn't
> converge, but I don't remember why, and failing to figure it out from
> looking at the vtu's. I guess I'll just try it again.
>
> Best regards,
> Lev Karatun.
>
> 2017-05-19 7:42 GMT-04:00 Glerum, A.C. (Anne) <A.C.Glerum at uu.nl>:
>
>> Hi all,
>>
>> Not sure if it is related, but this does remind me of the velocity
>> anomalies of the chunk geometry when the manifold was removed in the
>> interior of the domain after AMR (see #1361). But I don’t know how the
>> box geometry interacts with the manifold, if at all. Did you recently
>> update the dev versions of dealii or aspect Lev?
>>
>> Cheers,
>> Anne
>>
>> On 18 May 2017, at 21:40, Lev Karatun <lev.karatun at gmail.com> wrote:
>>
>> Hi Cedric,
>>
>> I'm using free slip for the bottom, free surface at the top, prescribed
>> in/outflow for the sides, and periodic boundary for front/back. I attached
>> the screenshots.
>>
>>
>> Best regards,
>> Lev Karatun.
>>
>> 2017-05-18 15:04 GMT-04:00 cedric thieulot <c.thieulot at uu.nl>:
>>
>>> Could you please show the viscosity (in log scale), the density and the
>>> temperature fields ?
>>> What kind of boundary conditions are you using ?
>>>
>>> Cedric.
>>>
>>>
>>> > On 18 May 2017, at 20:24, Lev Karatun <lev.karatun at gmail.com> wrote:
>>> >
>>> > Hi all,
>>> >
>>> > Lately, several of my simulations crashed with a non-convergence
>>> error, and in almost every case there was a velocity anomaly somewhere on
>>> the boundary between different refinement levels (see attached screenshot).
>>> If I turn off AMR, the problem goes away. I'm using an Iterated Stokes
>>> solver, the tolerance is currently 1e-6. I tried lowering the CFL number to
>>> 0.1, but it didn't help. I was wondering if anyone had similar problems in
>>> the past, and what would be the common approach to deal with them?
>>> >
>>> > Thanks in advance!
>>> >
>>> > Best regards,
>>> > Lev Karatun.
>>> > <mantle_vel_anomaly.png>____________________________________
>>> ___________
>>> > Aspect-devel mailing list
>>> > Aspect-devel at geodynamics.org
>>> > http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>>
>>> _______________________________________________
>>> Aspect-devel mailing list
>>> Aspect-devel at geodynamics.org
>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>
>>
>> <hanging_node_density.png><hanging_node_T.png><hanging_node_visc.png>
>> _______________________________________________
>> Aspect-devel mailing list
>> Aspect-devel at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>
>>
>>
>> _______________________________________________
>> 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/20170519/1cf392ba/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hanging_node_error_indicator.png
Type: image/png
Size: 33120 bytes
Desc: not available
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20170519/1cf392ba/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hanging_node_strainrate.png
Type: image/png
Size: 46465 bytes
Desc: not available
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20170519/1cf392ba/attachment-0003.png>