Need help in evaluating value of potential at given points inside the domain for mixed Neumann-Dirichlet Problem

41 views
Skip to first unread message

RAGHWENDRA KUMAR

unread,
Sep 30, 2019, 6:45:06 AM9/30/19
to bempp
Dear All,
       I am trying to solve a mixed Neumann-Dirichlet problem quite similar to the one given in Bempp documenation. Geometry that we have considered is slightly different, we have a hemiellipsoid of certain height(less than maximum domain length along z) sitting on z=0 plane. z=0 and z=50 planes are at fixed potential of 0 and 100 respectively. Potential of hemiellipsoid is also 0, rest of the boundary surfaces are  neumann boundaries with value 0.  I generate the geometry using "Freecad", mesh and mark surfaces using "gmsh", import in bempp script. Necessary changes are made in example script provided in bempp documentation to solve the problem. In the example script,  solution on the boundary surfaces have been plotted, I plot the same using "Paraview" and results looks reasonable. However when I try to evaluate the potential using Green's representation theorem, I get nan+ nanj  as the solution.

I have attached the script that I have adapted from example and the msh file of geometry with this post.

I shall highly appreciate any help in this regard.

Thanks
Raghwendra Kumar
 
hemi-ellip1.py
hemi-ellip.msh

Elwin

unread,
Sep 30, 2019, 7:56:00 AM9/30/19
to bempp
Dear Raghwendra Kumar,

If you specify the dual space in the evaluation of the new Dirichlet grid function, the NaN problem seems to be solved:
dirichlet_grid_fun_new = bm.GridFunction(dirichlet_space_dirichlet_segment,fun=dirichlet_data, dual_space=dual_dirichlet_space)

Best,
Elwin

RAGHWENDRA KUMAR

unread,
Sep 30, 2019, 8:41:22 AM9/30/19
to bempp
Dear Dr. Elwin,
             Thank you very much for prompt help. It is working fine now.

I noticed that for "neumann_grid_fun_new", it is working without dual_space while not in case  of "dirichlet_grid_fun_new".  I shall highly appreciate any insight for this observation.

With best regards,
Raghwendra
Reply all
Reply to author
Forward
0 new messages