Normal vectors in an NCI are flipped randomly (MechPDE, 3D)
Summary
I am working on a new implementation of a thinLayer formulation for a non-conforming interfaces for MechPDE. For this purpose, I need a normal vector in an integration point. But the orientation of a normal vector seems to be inconsistent as some of the vectors are flipped randomly.
Example Input or way to reproduce
I tested a 3D beam extension (fixed on the left end and negative pressure on the right end) with an NCI in the middle. The normal vectors shall be [-1, 0, 0]
, but some of them are flipped to [1, 0, 0]
. The merge request - !179. The test case files are attached as a comment.
What is the expected correct behavior?
Hardcoded [-1, 0, 0]
produces correct results. So, the normal vectors shall be consistent.
Possible fixes
Edited by Vladimir Filkin