[CIG-SHORT] Caldera ring structure in volcanic FE model

[Brad Aagaard]

Lucas,

I strongly recommend that you review the PyLith tutorial on using fault 
friction.
The slides and videos from the 2014 tutorial are available from
https://geodynamics.org/cig/events/calendar/2014-cdm-workshop/meeting-info/agenda/ 
and the slides from the 2016 tutorial are posted at 
https://wiki.geodynamics.org/_media/software:pylith:tutorials:cdm2016:pylithtutorial2016_friction.pdf

Slide 9 gives recommendations for solver tolerances:

[pylithapp.problem.interfaces.fault]
zero_tolerance = 1.0e-11

[pylithapp.petsc]
# Linear solver tolerances
ksp_rtol = 1.0e-20
ksp_atol = 1.0e-12
# Nonlinear solver tolerances
snes_rtol = 1.0e-20
snes_atol = 1.0e-10
# Set preconditioner for friction sensitivity solve
friction_pc_type = asm
friction_sub_pc_factor_shift_type = nonzero

See slide 8 for a discussion of the reasoning behind these values.

Regards,
Brad


On 08/11/2016 05:16 AM, Lucas Holden wrote:
> Hi
>
> thanks for your reply.
>
> The problem is a 3D problem. At present I have been using a model based
> on the volcano chamber/dike example in the Pylith wiki. I have
> added/used a cylinder structure above the magma chamber to represent the
> caldera ring structure.
>
> I use a dynamic fault interface with static friction (from the fault
> example in the manual section-7.8.7.2 Step10 and workshops). The caldera
> fault is modelled as a circular vertical ring (which I create using a
> vertical cylinder). The model runs, however there is no significant
> traction or slip occurring on the fault. I have also had to decrease the
> ksp_atol and snes_atol tolerance values to get the model to converge,
> however even with these values the convergence takes a while. These are
> at present:
>
> /ksp_rtol = 1.0e-10
> ksp_atol = 1.0e-8/
>
> /snes_rtol = 1.0e-10
> snes_atol = 1.0e-8/
>
> I have included my pylithapp.cfg and problem.cfg file as well as the
> current run log in this email.
>
> I will further investigate the mesh sizing example
> (examples/meshing/cubit_cellsize) to examine the influence of the mesh
> cell size around the structure, as suggested.
>
> Again, thanks for looking into my problem.
>
> Regards
>
> Lucas
>
>
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>