[CIG-SHORT] debugging rate-state quasi-static simulation

[Brad Aagaard]

On 1/15/18 11:58 AM, Ekaterina Bolotskaya wrote:
> Dear Brad,
>
> I removed the fault vertices from the boundary where I'm applying the Dirichlet BC as you suggested.
> Ideally, I would like to keep monitoring the slip on the fault after the whole fault surface has slipped (when the two halves of the block are sliding along each other). I have none of the vertices with both degrees of freedom fixed on either half.
> Could you tell me if Pylith would be able to converge after the two halves have separated.
> From what I see now it looks like the SNES solver stops converging when the fault slips completely.

With the current spontaneous rupture implementation, the nonlinear solve 
will converge very slowly once the entire through-going fault is 
slipping. The only remedy is to use small time steps and many SNES 
iterations.

Regards,
Brad