Repository navigation
Replies: 1 comment 2 replies
|
Hi Joroks! Let me first point you to the tracker issue for this #626 .
Definitively not a dead end! This is more a "we don't have enough hands right now" type of problem. I have started paving the way in this PR here #1188 which should simplify the next steps, as higher order elements are best implemented as tensor product elements (if possible). @AbdAlazezAhmed also has a prototype here #790 where, if I remember correctly, the DG case at least worked already. The missing key piece is the dof correction algorithm for face dofs in the dof handler here https://github.com/Ferrite-FEM/Ferrite.jl/blob/master/src/Dofs/DofHandler.jl#L709-L735 . A paper with one example how to implement it is referenced in the doc string. Right now the idea is that there is a "reference orientation" for each face and we simply reorder the dofs given in some orientation to match this reference orientation. This has the advantage that the number of cases to handle is quite small (i.e. 8 cases = 4 rotations and 2 reflrections). I am happy to help with further questions. Also, for performance reasons it can be helpful later to generate the dof index functions instead of splatting everything at runtime. |
Uh oh!
There was an error while loading. Please reload this page.
Hi!
I'm currently working on a problem that might benefit from higher order (GLL spaced) Lagrange interpolations so I tried to write my own implementation for these interpolations. I was relatively succesfull with this for lines and quadrilaterals. It should also work for Hexahedrons with interpolation orders up to 2.
Unfortunately, higher order interplations in 3D require more than one interiour facet dofs, which the DofHander currently doesn't support. Is this something that might change in the foreseeable future with maybe already some work in progress that I could check out or are higher order interpolations in 3D currently a dead end?
Here's my WIP for anyone interested:
All reactions