[CIG-SHORT] Elastoplastic material question

[Brad Aagaard]

Eric,

Can you send me a tarball of *all* of the input files, including the 
mesh? That will allow me to try to reproduce your error and test some 
things.

Another thing to try is to reduce the time step. If the increment in the 
load pushes the solution far outside the yield surface, it may have 
difficulty determining how to project back. I am skeptical that this 
issue is the problem, but it is relatively easy to try.

Brad


On 11/18/2013 06:30 PM, Eric Lindsey wrote:
> Yep, the models are converging. I don't think my solver is the best
> (still need to figure that out), but in this case they are working to
> the specified tolerance. When I decrease both the nonlinear/linear
> tolerances by a power of 10, there is no change in the results either.
>
> Lines from pylith's output just before the failure:
>
> ...
>   -- Solving equations.
>    0 SNES Function norm 8.855961010479e-05
>    Linear solve converged due to CONVERGED_ATOL iterations 111
>        Line search: Using full step: fnorm 8.855961010479e-05 gnorm
> 1.056268143324e-13
>    1 SNES Function norm 1.056268143324e-13
> Nonlinear solve converged due to CONVERGED_FNORM_ABS iterations 1
>   >> /home/class239/software/pylith/pylith-1.9.0-linux-x86_64/lib/python2.7/site-packages/pylith/problems/TimeDependent.py:200:run
>   -- timedependent(info)
>   -- Finishing advancing solution from t=3.15576e+07*s to t=6.31152e+07*s.
> Fatal error. Calling MPI_Abort() to abort PyLith application.
>
>
> Eric
>
> On Mon, Nov 18, 2013 at 5:09 PM, Brad Aagaard <baagaard at usgs.gov> wrote:
>> Eric,
>>
>> Have you verified convergence of the linear and nonlinear solvers up to the
>> point you get an error? If the solver doesn't converge, you will get garbage
>> that could show up as weird behavior.
>>
>> Brad
>>
>>
>>
>> On 11/18/2013 05:01 PM, Eric Lindsey wrote:
>>>
>>> I'm having trouble understanding the DruckerPragerPlaneStrain material.
>>> I'm
>>> using a simple homogeneous domain with no faults, and a simple Dirichlet
>>> BC
>>> imposing shear on the top/bottom. I've imposed an initial isotropic
>>> compression, then I add the shear displacement gradually until it should
>>> exceed the yield stress. On the sides I have a Neumann condition to
>>> maintain the normal stress; for the moment I just set the shear tractions
>>> to zero, but this value doesn't affect the results I'm getting. I expected
>>> to see uniform plastic shear throughout the domain, but instead get the
>>> message:
>>>
>>> RuntimeError: Infeasible stress state - cannot project back to yield
>>> surface.
>>>
>>> If I include "allow_tensile_yield = True", the model runs, but instead of
>>> homogeneous strain I'm getting the attached result. This is strange
>>> because
>>> at no point should the material be under absolute tension; the maximum
>>> shear stress in the elastic case is never larger than the magnitude of
>>> compressive stress. I think I am misunderstanding the setup of this
>>> material somehow; hopefully it's a simple error? Input files are attached,
>>> the elastic case works just fine. Any suggestions would be much
>>> appreciated.
>>>
>>> Thanks,
>>> Eric
>>>
>>> Relevant lines from the spatial databases:
>>>
>>>
>>>
>>> _______________________________________________
>>> CIG-SHORT mailing list
>>> CIG-SHORT at geodynamics.org
>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>>>
>>
>> _______________________________________________
>> CIG-SHORT mailing list
>> CIG-SHORT at geodynamics.org
>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>