Ina, Can you show the mesh on the plot? I am not sure that there is a problem, I think that the problem is only a visualization problem. The way visit and paraview work is to divide the rectangles in your mesh in triangles and then do a linear interpolation. You can really see on the coarse mesh that the cells
----- | | ----- are divided like this ----- | /| | / | ----- What you can do, is to use this function <https://dealii.org/current/doxygen/deal.II/classDataOut.html#a5eb51872b8736849bb7e8d2007fae086> to have better visualization of the coarse problem and compare it to the fine problem. Best, Bruno On Tuesday, March 26, 2019 at 9:18:58 AM UTC-4, Ina wrote: > > Hello everybody, > > having created a grid (let's call it coarse grid), copied and refined it > globally two times (let's call this one fine grid), I struggle with > transferring an interpolated function from the coarse to the fine grid. I > am using deal.ii 8.5.1. > > First I interpolated the function on the coarse grid using > VectorTools::interpolate, I will refer to it as coarse solution vector. > Then I followed two ideas: > > 1. Applying VectorTools::interpolate_to_different_mesh. This works well > for simple functions like f(x,y)=x or f(x,y)=x*y but unfortunately not for > the CurvedRidges function from step-13 (see attachment). I don't get why it > does not work. Looking into the code of interpolate_to_different_mesh, I > saw that it essentially calls DoFCellAccessor::get_interpolated_dof_values > and set_dof_values_by_interpolation. > > 2. Create an InterGridMap from the coarse to the fine grid, use it to call > DoFTools::compute_intergrid_transfer_representation. Unfortunately, the > documentation of this function does not fit the actual code (any more). The > output data type seems to be changed from a matrix to a > std::vector<std::map<.,.>> (I also think that in the comments of the code > "row" and "column" are mixed up). Since it is not a matrix any more, a > vector can not simply be applied using multiplication. By looping over the > coarse and fine solution vector, identifying the contribution weights using > the transfer representation and writing the respective entry to the fine > solution vector, the coarse dof entries can be identified with the > corresponding fine grid entries. Now the values still need to be > interpolated. According to the documentation the output of > commpute_intergrid_transfer_representation can be used to construct a > corresponding sparse transfer matrix. Unfortunately, I did not find such a > function that accepts a std::vector<std::map<.,.>> in the documentation. > Did I just miss it or doesn't it exist? > > Do you know why my approach 1 does not work? Do you have any further ideas > on how to transfer an interpolated function between meshes? > > I would greatly appreciate your help. > Ina > -- 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 [email protected]. For more options, visit https://groups.google.com/d/optout.
