[CIG-SHORT] Pylith troubleshooting

[Brad Aagaard]

Marcello,

I have never tried to push the fault open and then closed again via a 
traction perturbation. I have a cyclic loading test in which I move the 
boundaries of the domain to open/close the fault. It works as designed. 
That is, the fault is traction free when open and friction returns when 
the fault is closed.

Your traction perturbation seems quite extreme. I think you are trying 
to model the exploding wire used to nucleate the slip. The dynamic 
elastic deformation surrounding the nucleation zone may help drive slip 
so that you don't need to force the fault into tension via the traction 
perturbation. A smaller amplitude perturbation might be sufficient. Have 
you investigated this?

Nevertheless, if you try to drive the fault closed again with a negative 
traction perturbation you should see a difference when open_free_surface 
is set to False (default is True).

If open_free_surface = False, then when the fault is open the traction 
perturbation will still be applied. However, whether it is sufficient to 
close the fault depends on the dynamic elastic deformation.

If open_free_surface = True (default), then when the fault is open the 
traction perturbation is not applied. If the fault is open, the fault is 
traction free.

Regards,
Brad


On 2/10/15 1:45 PM, Marcello Gori wrote:
> Dear Brad,
>
> Thanks a lot for your previous help. Now the code works fine and gives
> me nice results.
>
> On a different note, I cannot understand why it behaves in the following
> way with the traction_perturbation condition:
>
> I have a geometry like the one in the figure below, for which, in order
> to prescribe the load P, I specify an initial compressive stress of
> sig_y = -P/A inside in the entire bulk.
> The resolved normal tractions on the fault due to this initial stress
> result -P/A*cos(alpha)^2.
>
> Then I apply an opening traction condition on the fault using the
> db_change and th_change (let's suppose a ramp-up, plateau, ramp-down for
> both profiles (see figure below)).
>
> The spacial ramp (db_change) is equal an opposite to the resolved normal
> stress (i.e. +P/A*cos(alpha)^2, while the temporal ramp (th_change) varies.
> Therefore, for th_change values <= 1 there is no opening, and for values
>  > 1 opening will occur.
>
> The strange behavior is that I impose a ramp-up, plateau, ramp-down
> profile in time: for th_change values <= 1, this profile is perfectly
> followed by Pylith; conversely, for th_change values > 1, this profile
> is followed only for the ramp-up, then Pylith keeps one value for the
> whole duration of the simulation (regardless of the shape and the actual
> value of the profile), and this value is equal to 1 (opening condition).
> Is there an opening condition I am not aware of? Also, how can I force
> Pylith to follow that profile and then recover back the closing and
> traction condition specified in the th_change file?
>
> I hope this was clear, but I can send you output files to clarify, if
> needed.
> I have tried to set open_free_surface to both True and False, but it
> doesn't change this behavior.
>
> Thank you very much for your help.
> Best wishes,
>
> -Marcello
>
> Inline image 1
> Inline image 2