Hello everyone, I am using deal.ii to solve a 3D solid mechanics problem. My code uses Parallel::Distributed::Triangulation and Petsc Wrappers.
Until now, the code does not support AMR, however, I want to implement it in the code. For the last couple of days, I am looking at the example codes (esp. 31) that talk about how to do adaptive mesh refinement and interpolate the solution to the new mesh. I have read several previous posts in this group as well. Before actually starting to code, I wanted to ask a few questions to make sure I understand the concept and should be able to debug when the need arises. (I think this is very closely related to serialize/deserialize) The adaptive mesh refinement works as follows: 1. Once the solution is known on a coarse mesh, and mesh is refined locally in some regions, SolutionTransfer class is used to get the solution interpolated on the new mesh. 2. In order to have gauss point history to be interpolated to the new mesh, I can make a new vector which is numbered as per the dif handler of discontinuous Galerkin. I can then store the gauss point history in this vector. This vector can then be interpolated to the new mesh and its value at the Gauss points will give the new point history. 3. Constraints related to hanging nodes have to be taken care. This will actually transfer the solution and point history from the old mesh to the new mesh. I also have 2 questions now: 1. Apart from the transfer of these things (like explained above), is there anything else one needs to keep in mind while working on AMR? 2. After repartition, there can be 3 cases that *a)* a new cell (previously belonging to the different processor) becomes a part as a result of repartition. *b)* there are newly refined cells with parent cell belonging to same processors (pure refinement) *c)* there are newly refined cells with parent cell belonging to same processors (refinement + repartition). In case of no repartition, are material id and boundary conditions flags transferred to newly generated cells? In other words, as long as no motion of cells from one processor to other takes place, the data is inherited from parent cells. Is this right? In case of repartition, are the material id and boundary conditions transferred along with these cells (when they move)? The link https://www.dealii.org/8.5.0/doxygen/deal.II/classparallel_1_1distributed_1_1Triangulation.html says that when cells are moved from one processor to other, this information is not carried. Thanks. -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
