[CIG-SHORT] Dynamic rupture on realistic mesh

[Brad Aagaard]

Huihui,

In a 3-D simulation with a 2-D fault surface and spontaneous rupture, 
slip occurs in the direction of the current shear traction vector. Even 
if you apply initial shear tractions in only one direction, the dynamic 
shear tractions may be large enough to cause slip to be in a slightly 
different direction. Dynamic stress changes are affected by spatial 
variations in elastic properties, nonplanar fault geometry, and 
discretization error associated with coarse or distorted cells.

The first thing I would do is to run a problem with the same parameters 
and a planar fault. You should check the mesh quality, 
nondimensionalization scales, and how well resolved the solution is by 
running at a few different resolutions.

Once the planar fault produces good results, then start increasing the 
roughness (nonplanar geometry) of the fault. Start with a very smooth 
fault and then gradually work your way to the desired roughness. If you 
have significant curvature in the fault surface, you may need to 
increase the resolution of your mesh in those areas. You should be able 
to identify those areas with careful analysis of the solution as you 
increase the roughness and evaluate multiple resolutions. As we point 
out in our tutorials, you should consider the uncertainty and resolution 
of the observations that you are trying to model when deciding on the 
length scales to include in your model. For example, we know the 
topography of Earth's surface very well, but we usually don't need a 
high resolution discretization of topography in our models due to much 
larger uncertainties in geometry and properties at depth.

Regards,
Brad



On 11/08/2017 10:02 PM, Huihui Weng wrote:
> Hi Brad,
> 
> I have rerun this case by single computer core. In this situation, the 
> progress will abort with error: RuntimeError: WARNING! Fault opening 
> with nonzero traction., v_fault: 7990, opening: 1.0026e-07, normal 
> traction: -0.00370101. Please see the attached file for detailed log.
> 
> The most important problem I concern about is why the normal and strike 
> slips always accompany with the dip slip in the curved fault, even only 
> assigning the shear stress at the dip direction. How can I inhibit these 
> unnecessary slip? Could I do it by adjust the parameters of PETSc? Thanks.
> 
> Best,
> Huihui
>