[dune-pdelab] Hanging Nodes

[schauer]

Dear PDELabers,

I try to implement hanging nodes together with Dirichlet constraints. I 
acted as in the poisson example of the pdelab howto. 

But now some things appear strange to me:

- The transformation of a vector with 
  gos.backtransform 

  as well as
  
  gos.forwardtransform
  
  only seem to multiply the hanging node entry of the input-vector 
  with a value from the constraints container without any kind of
  interpolation with the neighbouring nodes - so the transformation
  matrix is diagonal and in my case the backward transformation after a
  forward transformation do not give the original vector but in my case 
  these transformations even seem to be identical. 

I then looked at the constraints container with the verbosity flag in
the constraints call, and also saw that all the hanging nodes had just
one entry with the same line as row index, e.g. like:

266: (266,0.25) 
267: (267,0.25) 
268: (268,0.25) 
275: (275,0.5) 
276: (276,0.5) 
278: (278,0.5) 

which also consists of only one value in every row (somehow I supposed
the neighbouring, non-constrained nodes should show up here). 

The row entries of the resulting matrix at a hanging node position were
trivial, which seems to be fine to me, but I don't think that the other
rows were changed correctly (though I didn't verify).

I suppose these things can also be verified on the hanging nodes example
poisson.cc of the pdelab-howto.


So is the way the constraints container gets filled for the hanging
nodes wrong - or am I not correct with some of my assumptions ? 

Thanks

Volker
 




-- 

_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/

Volker Schauer (Dipl.Phys)

Institute of Applied Mechanics (Civil Engineering), Chair 1
University of Stuttgart
Pfaffenwaldring 7
D-70569 Stuttgart
GERMANY

Phone ++49 (0)711-68560044
email: schauer at mechbau.uni-stuttgart.de