dolfin team mailing list archive

[Patrick Riesen <priesen@xxxxxxxxxxxxxxx>]

Anders Logg wrote:
> On Mon, Apr 19, 2010 at 10:39:26AM +0200, Patrick Riesen wrote:
> > hello,
> >
> > I possibly want to implement a free surface problem with my dolfin
> > code using the dolfin ALE functionality. dolfin can do the automatic
> > updating of the mesh nodes or else shift the nodes by a
> > user-supplied displacement function.
> >
> > in the automatic updating, the vertex-displacements are
> > 'method-supplied' and not accessible, but i need to have them in
> > some displacement function  in order to compute correct convective
> > velocities at all theeuler/ale/lagrange dofs.
> >
> > i can get the displacements at the mesh vertices from comparing old
> > (not yet moved) and new (moved) mesh, but what's a good way to
> > generate a displacement function in a desired functionspace out of
> > those vertex-values?
>
> What is you input?
>
> Is it displacement on the boundary? Then you can just send in the
> displaced boundary as argument to mesh.move().
i define how the free boundary part will be handled/displaced, and then 
i use dolfin ALE to automatically adjust the interior nodes.

but then don't i need the displacements applied by dolfin at the 
interior nodes to keep track of the relative displacement between mesh 
and fluid everywhere in the domain (and my equations)?

> Or is it a vector-valued function (perhaps obtained by solving some
> PDE for the displacement). 

Then just send in that function to
> mesh.move().
>
> If you already have the displacement of the mesh, why do you need the
> ALE functionality? If the mesh is already displaced, then nothing
> needs to happen.
>
> --
> Anders