On Wed, 17 Sep 2014, Harshad Sahasrabudhe wrote:

> I'm getting a few spurious eigenvalues from an EigenSystem when I do
> adaptive refinement using KellyErrorEstimator. Has anyone faced this
> problem before?

If you use the asymmetric_constraint_rows option (which is on by
default when you apply constraints manually), then you'll be adding
spurious eigenvalues to the matrix in hard to predict ways.

If you don't use that option (which is off by default when you use
FEMSystem), then you'll be adding an eigenvalue of 1 to your matrix
with a multiplicity equal to your number of constrained dofs.

If you're using EigenSystem you can probably switch to
CondensedEigenSystem to get rid of the spurious values.

If you're using FEMSystem+EigenTimeSolver you're currently out of
luck, but it shouldn't be hard to adapt the relevant code from
condensed_eigen_system.C into eigen_time_solver.C; we'd love a patch.
---
Roy

------------------------------------------------------------------------------
Want excitement?
Manually upgrade your production database.
When you want reliability, choose Perforce
Perforce version control. Predictably reliable.
http://pubads.g.doubleclick.net/gampad/clk?id=157508191&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to