[CIG-SHORT] Dynamic rupture on realistic mesh

[Brad Aagaard]

Huiui,

My guess is that the "freeze" is occurring when PyLith has encountered 
an error and at least one process has aborted, but the MPI 
implementation is not detecting it. When this has happened to me, I will 
see there errors sent to stdout or stderr, which may be on a local 
compute node if you are on a cluster (this is MPI/machine dependent).

I will look deeper into your description when I get a chance.

Regards,
Brad


On 11/8/17 4:48 AM, Huihui Weng wrote:
> Hi Brad,
>
> When I try to simulate dynamic rupture with realistic mesh, I have a
> similar problem with the one Hongfeng asked several years ago. But in my
> case, the normal stress is large enough to hold the fault (attached
> pictures).
>
> For test, I let the dip shear stress at one node reaching the fault
> strength, but I find it also has slips at both normal and strike
> directions (as the attached pictures show, information at left bottom),
> i.e., dip slip ~ 0.05m, strike slip ~ 0.005m, normal slip ~ 8e-5m (fault
> open with normal stress large enough!). The simulating progress of this
> case will freeze at some time step without any errors, and the freezing
> time step depends on the parameter of zero_tolerance (e.g., 1.0e-7).
> Once the progress freezes, I need to kill it by hand by Linux commands.
>
> Since only the dip shear stress and normal stress are not zeros in this
> case, how could it lead to strike and normal slip? It does seem it is
> the abnormal normal slip that “freezes" the simulating progress. How can
> eliminate the slip at the normal direction for such realistic fault? Thanks.
>
> Best,
> Huihui
>