[CIG-SHORT] Pylith question

[Demian Gomez]

Hi Matt,

1) It would be nice to see the full, uniform sphere test with comparison to
analytic solution
See attached figures. This test was done using a uniform material sphere
with slip on a single node. Since Pollitz' analytic solution uses
rectangular uniform slip patches, I estimate (by trial and error) the scale
size and amount of slip of a patch that has the same orientation as the
finite element node that I use in Pylith. The agreement in the mid and far
field is excellent, but in the near field you can clearly notice that there
are some differences, which I want to double check with and without
gravity. In general (except for some near field stations), the difference
is below 2 mm.

2) So you are saying that you only see a problem with gravity when you use
layers?
No. Sorry if I didn't express myself correctly. The problem
(non-convergence) occurs no matter what model I use. With or without layers.

3) If these problems are time dependent, it is essential to look at
convergence in the time step, since you could have instability when turning
on gravity.
These simulations are not time dependent. I am simulating quasi static
coseismic deformation.

What should be the next step?

Cheers,
Demián

On Thu, Jan 14, 2016 at 12:12 PM, Matthew Knepley <knepley at ci.uchicago.edu>
wrote:

> On Thu, Jan 14, 2016 at 10:59 AM, Demian Gomez <demiang at gmail.com> wrote:
>
>> Hi Matt,
>>
>> Thanks for your quick answer. I can definitely do that. I jumped a little
>> bit ahead because I'm using the analytic solution from Fred Pollitz
>> (static1d), which handles sphericity and layers. My model's far-field
>> solution looks just like the analytic solution from Pollitz, but the near
>> field has some discrepancies. I'm not exactly sure what the problem is, but
>> because Pollitz' code includes gravity, to compare "apples with apples" I
>> wanted to activate gravity in Pylith too.
>>
>> I can take a step back and make my sphere uniform. Would it be OK to use
>> the same mesh and just apply the same spatial database to all layers? Or
>> should I make a model with no layers at all?
>>
>
> 1) It would be nice to see the full, uniform sphere test with comparison
> to analytic solution
>
> 2) So you are saying that you only see a problem with gravity when you use
> layers?
>
> 3) If these problems are time dependent, it is essential to look at
> convergence in the time step, since you could have instability when turning
> on gravity.
>
>   Thanks,
>
>      Matt
>
>
>> Thanks again,
>> Demián
>>
>> PS: during my tests, using a uniform sphere produces the same result when
>> I activate gravity
>>
>> On Thu, Jan 14, 2016 at 8:31 AM, Matthew Knepley <knepley at ci.uchicago.edu
>> > wrote:
>>
>>> On Wed, Jan 13, 2016 at 10:43 PM, Demian Gomez <demiang at gmail.com>
>>> wrote:
>>>
>>>> Dear Matt,
>>>>
>>>> I'm Demián Gómez from The University of Memphis and I work with Eunseo
>>>> Choi. We briefly met in 2014 during the Pylith workshop at Stanford. Back
>>>> then, I was just starting to use Pylith with a full spherical model (not
>>>> just a section of the sphere) that you may remember.
>>>>
>>>> I'm bothering you with a question because I was unable to find a
>>>> solution to my problem, so I decided it was time to ask you. I've been
>>>> trying to add gravity to my spherical model, which I've been using to
>>>> obtain the coseismic deformation from the Maule earthquake. I activate
>>>> gravity as shown in the manual (gravity_field =
>>>> spatialdata.spatialdb.GravityField) with the coordinate system set
>>>> to coordsys = spatialdata.geocoords.CSGeo in the pylithapp.cfg (version
>>>> 2.1.0). My spatial databases are configured to have is-geocentric = true
>>>> and ellipsoid = sphere to specify that my mesh of Cartesian XYZ coordinates
>>>> refer to the geocenter (section C.2.1.2 of the manual).
>>>>
>>>> Everything works fine until I activate gravity. When I turn gravity on,
>>>> the problem never converges and the residual norm shown during the
>>>> integration process is very large (in cases without gravity it starts at
>>>> ~10^-4 and converges at ~10^-12, with gravity it is ~ 10^2 and it never
>>>> decreases!). I don't know if it's important, but without gravity my model
>>>> converges in ~250 iterations.
>>>>
>>>> I am not sure if this is a configuration problem or something that
>>>> Pylith cannot handle. If you could please point me in the right direction
>>>> to correct this problem, I'd really appreciate it. Maybe there is a way to
>>>> specify a spatialdb with the direction of the gravity field? I read the
>>>> documentation but I could not figure it out how to do it.
>>>>
>>>> If you need more information (I could send you the complete model if
>>>> needed), or if I should direct this question to somebody else, please let
>>>> me know. I've attached the config files and one spatialdb just in case you
>>>> want to take a look at them.
>>>>
>>>
>>> Hi Demian,
>>>
>>> With something like this, I think the right thing to do is get as simple
>>> as possible first.
>>>
>>> So you are solving a whole sphere problem, which we never do in our
>>> examples. Thus we should
>>> make a simple example for this that we completely understand. How about
>>> making a uniform sphere
>>> with gravity directed toward the center. This is analytically tractable.
>>> If you can set this up, we will
>>> incorporate it into the PyLith test examples, and work out an analytic
>>> solution.
>>>
>>> If that works, we can start adding back the complications in your model.
>>>
>>>   Thanks,
>>>
>>>      Matt
>>>
>>>
>>>> Thank you for any help you can provide me,
>>>> Demián Gómez
>>>>
>>>> --
>>>> *Demián D. Gómez*
>>>> Graduate Research Assistant & PhD Candidate
>>>> Center for Earthquake Research and Information
>>>> 3890 Central Ave, Memphis TN 38152
>>>> Tel: +1 (901) 678-4809
>>>> Cell: +1 (901) 900-7324
>>>>
>>>
>>>
>>>
>>> --
>>> What most experimenters take for granted before they begin their
>>> experiments is infinitely more interesting than any results to which their
>>> experiments lead.
>>> -- Norbert Wiener
>>>
>>
>>
>
>
> --
> What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> -- Norbert Wiener
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-short/attachments/20160115/a520ca91/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: map_gps.png
Type: image/png
Size: 200819 bytes
Desc: not available
URL: <http://lists.geodynamics.org/pipermail/cig-short/attachments/20160115/a520ca91/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: differences_NEU.png
Type: image/png
Size: 11267 bytes
Desc: not available
URL: <http://lists.geodynamics.org/pipermail/cig-short/attachments/20160115/a520ca91/attachment-0003.png>