Hi Jack, The student who was working on isoparametric mappings last summer didn't get very far, however we were very fortunate to host Martin Alnæs for six weeks this spring. We designed, and he implemented, the UFL layer changes needed to get isoparametric in. We still have some legwork to do in our FFC branch to get those changes into Firedrake and I believe the equivalent work for Dolfin is still underway, but there is progress and we hope to have isoparametric by the summer.
On the shell subject, Firedrake now supports simulation on extruded shells formed of triangular prism elements. I'm not sure if that helps you but you're welcome to give it a try. Regards, David On 16 April 2014 10:39, Jack HALE <[email protected]> wrote: > Hi Corrado! > > Sorry for the slow reply on here, I know we have discussed this > privately, but of course this is the best place for discussion. > > 1. First half of the presentation; My understanding of the current > 'PDE on Manifold' functionality in FEniCS is that the weak form cannot > include terms relating to the geometry of the manifold. i.e. it would > be natural to have terms such as the fundamental form expressed > through UFL which you could then define the shell model. > > I have seen someone discuss this idea before here: > http://www.mail-archive.com/[email protected]/msg08932.html > > albeit in the context of isoparametric mappings. I think though, that > isoparametric mapping is just relating R^3->E^3 and the shell concept > is relating R^2->E^3, the efforts towards shell models should work > within bringing isoparametric mappings to FEniCS. > > @David Ham: I remember David Ham discussed with me that he had a > student working isoparametric mappings, did anything come of it? > > 2. Second half of the presentation; local projections. As you can see > I have done some simple local projections at the linear algebra level > (ie. post assembly), but I do not think this is a suitable path for > implementing the MITC operators which are significantly more > complicated. One initial option would be to do the full mixed problem, > at the expense of engendering extra unknowns. Also you suggested in > our private email that we could do these local projections using a > custom C++ kernel/assembly routine. > > I can see there are still some problems with the RT elements on > manifolds, it would be important for this functionality to work first: > > http://fenicsproject.org/pipermail/fenics/2014-March/001340.html > > And only two threads up from this one, this discussion seems pertinent: > > http://fenicsproject.org/pipermail/fenics/2014-March/001371.html > > Another option is that we avoid this second piece of functionality and > go with trying to get DG-Koiter shell models working first which work > which are rotation-free and use standard element constructions. > > @Garth Wells: I know this is something Garth Wells is an expert on so > perhaps it is the best path forward for now? > > 3. Generality. So I know a lot about shells, but not about other PDEs > on manifolds. I remember Douglas Arnold mentioned that any approach > implemented in FEniCS > should be as general as possible. Any comments on this? > > Kind regards, > ----- > Dr. Jack S. Hale > > Research Associate > > University of Luxembourg > Campus Kirchberg G005 > Phone +352 44 66 44 5236 > [email protected] > > Latest publications and conferences: http://goo.gl/rNiISG > ORCID: http://orcid.org/0000-0001-7216-861X > Google Scholar: > http://scholar.google.com/citations?user=Fx9lQ7MAAAAJ&hl=de > _______________________________________________ > fenics mailing list > [email protected] > http://fenicsproject.org/mailman/listinfo/fenics > -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham
_______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
