[CIG-SHORT] Zero initial tractions

[Brad Aagaard]

On 03/12/2015 09:40 AM, Ge Li, Mr wrote:
> Hi all ,
>
> I had some weird stuff and questions about zero initial tractions and fault opening in the output.
> 1.
> All traction values are at t=0.0*s are zero, e.g. benchmark results
> of tpv205, point (0.0, 7.5),h-shear-stress, although initial horizontal shear traction prescribed here is 81.6MPa
> Static friction is 81.24MPa and dynamic friction is 63MPa. Time step is chosen as 0.05s in this example
> and h-shear-stress is 81.10MPa,slightly smaller than 81.24MPa.
> So first question is: Why traction value at t=0.0*s is not equal to prescribe initial tractions?
> Although it seems that this problem doesn't affect following stress evolution and yield similar results to other codes, e.g. eqsim, I am still curious about it and want to figure out why.

The zero tractions at t=0 is an artifact of how we do the output and how 
the simulation starts up with explicit time stepping.

> 2.
> I did some tests by setting time step equal 0.02,0.01,0.005 and 0.002 respectively and found identical patterns
> that initial zero followed by a traction jump after one time step. Interestingly, h-shear-stress approaches 81.24MPa as time
> step decreases and reaches 81.24 at 0.002*s.
> So second question is: It seems that there is a minimum time step above which complete evolution cannot be captured. But how to determine it?

Just in the same way there is discretization error in space, there is 
discretization error in time as well. The central difference method, 
which we use for explicit time stepping, is second order in time.

> 3.
>   Under 'open free surface = true' case , I find some points where fault opens with none-zero tractions  These points are selected using Paraview find data function(slip[:,2]>10e-10 & abs(traction[:,2])>10e-10).
> This  is a link my .cfg file and output.:
>
> https://www.dropbox.com/sh/8s4m448qy8i8id4/AAACD73Kc8tYSYs___5XOoKEa?dl=0
>
> There is also a screenshot of selected point lists from Paraview.
>
> This phenomenon  should be impossible considering a 'True' case! If there is something about output should be noticed?
>
> Looking forward to your reply! Thanks !

In the "True" folder, for the fault.h5 and fault.xmf files, I don't see 
any points with slip[:,2] > 0. I do see a few points with traction[:,2] 
 > 0, but the tractions are on the order of a 10 kPa while the shear 
tractions are on the order of 10 MPa. This seems reasonable given your 
nonplanar geometry. I don't quite follow what you are asking or what you 
think the problem is.

Regards,
Brad