dolfin team mailing list archive

[Jed Brown <jed@xxxxxxxx>]

On Tue 2009-03-17 11:41, Matthew Knepley wrote:
> Regularization of the viscosity is a little more tricky. Can you show that this
> solution converges to the true solution?

In this case, regularization is actually physical and without it
viscosity may become infinite.  In many cases the regularization can be
made small enough to support viscosity variation of nearly
machine_epsilon^((1-n)/n), the best you can expect given strain rates
with dynamic range of 1/epsilon.

In my experience, a line search can significantly reduce the number of
continuation steps necessary.  Enough to more than make up for the
greater number of Newton iterations per continuation step.

Jed

Attachment: pgpIuDYTGHbuW.pgp
Description: PGP signature