On 01/30/2012 02:26 PM, Torsten Bronger wrote:
I get the warning
warning: singular matrix zeroing nan and inf entries... warning: (almost) singular matrix! (estimated cond. number: inf)
because I have nodes in the mesh that don't belong to any element. I could get rid of them, however, this would make my postprocessing algorithm more complicated. Can I safely ignore this message?
You should not ignore the warning. If the solution seems ok, it might be due to robustness of umfpack - it's not to be expected with another linear solver.
If you do not wish to delete the nodes, try defining a region containing only them, and use the region to set zero Dirichlet boundary condition (ebcs) there. (Never tried that though - if it fails, chime in with the error message - it should work.)