[CIG-SHORT] How to mesh the 2-D model into small size

[Brad Aagaard]

On 11/27/2017 11:26 PM, xu bei wrote:
> Dear Charles
> 
> I constructed a 2-D mesh model with two faults. I want to see whether
> the dynamic slip can jump from one fault to the other fault. At
> first, I set the grid size of 100 m and the fault zero tolerance 
> value 1.0e-10. But some nodes open with nonzero traction (see
> attached log file). Idecrease the grid size and want to eliminatethe
> problem. But PyLith will estimate a stable time step based on the
> minimum grid length of your mesh. Although Idesigned a grid size of
> 50 m or smaller in CUBIT, the actual grid size on certain cells may
> be extremely small that depends on the mesh quality.
As I mentioned previously, if you decrease the cell size, then the time 
step for stable integration (CFL condition) is reduced by the 
corresponding amount. The remedy is to simply reduce the time step as 
you use a finer mesh.

  Then I increase
> the size to 60 m or 80 m. The meshquality is high, but some nodes
> still open with nonzero traction (see attached log file).

Are you resolving the cohesive zone with 3-5 cells at this resolution. 
If not, then you need either a finer mesh or need to adjust your 
friction model parameters to increase the size of the cohesive zone.

  I also
> increase the fault zero tolerance value to 1.0e-7. But I do not know
> whether increase the fault zero tolerance value will increase the
> numerical noise.

  And for a grid size of 60 m, although I increase the
> fault zero tolerance value to 1.0e-7, Pylith still report that some
> nodes open with nonzero traction (see attached log file).

In v2.2.1, the fault has a zero_tolerance and a zero_tolerance_normal. 
You should set both as the zero_tolerance_normal value is still the 
default value of 1.0e-10. Sorry for the confusion.

I increased the zero_tolerance_normal value to 1.0e-5. The fault still 
opens and it looks like the rupture is supershear. When the supershear 
rupture hits the large fault bend, the fault is tending to open. So I 
would suggest adjusting the friction and stress parameters so that you 
have a subshear rupture. You may still need a finer mesh and/or a larger 
zero_tolerance_normal for the rupture to make it around your large bend 
without opening.

It is easy to overlook the default values in the manual. This is why we 
strongly recommend using the PyLith Parameter Viewer when debugging; it 
shows all of the values and where they are coming from.

Regards,
Brad