On Mon, 13 Nov 2006, Michael Povolotskyi wrote:
> I'd like to solve a linear system on a 2D mesh that lies in the (y,z)
> plane, not in the (x,y) plane.
> Is this possible?
It's supposed to be possible - I don't know of any of the developers
solving such problems, though, so don't trust the results without
doing some verification studies.
> Will the KellyErrorEstimator work correctly in this case?
It should. Kelly just knows about gradients and normals, and
fe_boundary.C is supposed to calculate both correcly for 2D elements
living in 3D space. If you were solving on a bent surface you might
have some problems with overrefinement into surface edges, but a flat
plane shouldn't be any trouble.
As an aside, though, I just noticed that we're not calculating normals
correctly for 1D elements living in 3D space; I'll see if I can get
that fixed this week.