search

SCOREC / core

parallel finite element unstructured meshes
Other
184 stars 62 forks source link

Issue with recovered field in SPR #349

Open mortezah opened 3 years ago

mortezah commented 3 years ago

There seems to be a bug inside SPR routines, that causes the successively adapted meshes to not converge for some examples. For example, attached is a picture of adaptive-solve (with MFEM and PUMI) for a linear elasticity problem on the upright model. In this picture going from the 1st adapted mesh (middle mesh in the top row) to the 2nd adapted mesh (right mesh in the top row), the mesh gets coarsened a lot.

pumi_mfem_successive_adapted_meshes_upright_model

The following picture shows the recovered fields (computed inside SPR routines) on the same meshes. Note the extremely high value of the recovered field on the 1st adapted mesh (middle mesh in the top row).

pumi_mfem_successive_recovered_field_upright_model

cwsmith commented 2 years ago

I think this may have been an off-by-one error. https://github.com/SCOREC/mfem/commit/2d055eb0793670cf4bad3eca3a47e6967d9b783c

mortezah commented 2 years ago

That could be the case.