dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Mar 22, 2011 at 10:29:07AM -0000, Bento wrote:
> Question #149906 on DOLFIN changed:
> https://answers.launchpad.net/dolfin/+question/149906
>
> Bento posted a new comment:
> I am experimenting with SubMesh. I created a submesh over the domain
> where I want to solve the Stokes equation. But how can I expand the
> solution to the bigger mesh? I set allow_extrapolation to True and then
> projected and interpolated the solution to the bigger mesh. However the
> extrapolation process takes much longer than solving the problem. Is
> there a better way for this?

It might be helpful to look at the FSI solver in CBC.Solve which you
can get by

  bzr clone lp:cbc.solve

Check the code in sandbox/fsi. It uses SubMesh to define different
domains and uses two different approaches: first iterating between
solutions on two different meshes (F) and (S) but also having a full
representation of the solution on the union of the meshes (used in the
so-called dual problem) in combination with the ident_zeros function
in the matrix class to "remove" constrained degrees of freedom.

--
Anders