[petsc-users] Pseudo time stepping

[Jed Brown]

Dharmendar Reddy <dharmareddy84 at gmail.com> writes:

> Hello,
> I need to slove a nonlinear problem :
> F(x)= 0  with x = xd specified on the boundary. I have setup an snes based
> solver. I need to modify it to handle the situation described below.
>
> The bc value xd  takes a set of values
> xd = {x1, x2 ......., xn}, xj > xi for j > i .
>
> Parametrising xd(t)= x1 + (xn - x1) * t  such that by selcting t between 0
> and 1 I get all the xd values mentioned above.
>
> I want the solver to automatically increase or decrease the t based on
> convergence criteria ( i.e., go back to earlier converged step and try
> again with a smaller step, ofcourse report divergence if the step size is
> smaller than a specified minimum step size) and solve system of equations.

See TSPSEUDO, which uses residual-based step control.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20140303/a200b313/attachment.pgp>