Thanks Brad, yes, I know that one too. Aslo (and sometimes) when over the edge of the geometry but always when overlapping a surface/volume region.
In this case none of these. Had me scratchng my head as the model run (convergence of contacts took all weekend), deformed shape, displacements and stress patterns looked intuitive, but the forces simply did not add up.
... It turns out it's the 'split surfaces' of the contact between the components. A surface region that's hidden in the prep. The overlap was slight. I have tweaked the test diameter and the residual norms at the contact seem well behaved and monotonic this time.
It is peculiar how the s/w interpretted this overlap to give size alternating and some missing test diameters.
Presumably the force calculation is confused by the elements 'existing' in multiple regions and the s/w doesn't understand how to add 'em up.
Do you know if we switch off these test diameters somehow? 99% of the time we use contact regions and therefore the test diameters are not of any use.
For others out there, this may be helpful ...
symptoms can include lengthy meshing with complaints including overlapping geometry with the diagnostics putting up little red circles or purple crosses in the model at locationss that seem to bear absolutely no relation with the actual location of the problem. The clue here is 'overlapping geometry', the s/w is right, but you can't see it as the test diameter surface region is not created by the user. The only way to look for it is to mesh one component at a time so you can see the interface surfaces and the automatically created annular region ... and then you have to remember that it could be the split surfaces.
Subtly misaligned fasteners can give similar symptoms. The misalignment can be fractional (often as a result of 2 components having had their holes defined outwith a skeleton, or having lost their external references at some point ...) Gross misalignment means the fastener cannot be created. Mechanica 'ignores' subtle misalignment and then gives you a hard time when meshing followed by daft answers if it runs.
.
