On the representation of normal displacements of curved contacts

3DEC
1

Hi everyone !

That will be two posts in a row for me today :slight_smile:.

We study the impacts on considering irregular contacts in 3DEC.

Thanks to the block merge commands, we can import meshes that were created using dedicated tools (e.g., Salome).

Modeling a really simple scenario (single elastoplastic joint in an elastic cube; gravity; oedometric boundary conditions; and purely mechanical), we compared a model with a flat contact (3DEC created) and a very slightly bended contact (Salome created then imported into 3DEC). The bended plane is really close to the flat one:

image
image505Ă—554 4.2 KB

With similar mesh precision, results do differ although the geometries are close:

image
image1319Ă—681 221 KB

However, representation of normal displacements/stresses bring strange contours with apparent discontinuities, as if there were multipe results per gridpoint.

image
image1332Ă—671 225 KB

Having this kind of multiplicity on normal displacement and not shear ones might result from the projection on the (slightly) different normal vectors of subcontacts.

My question is: is this an effect of visualisation or is there actually “multiple” results per subcontact?

I am not sure how merge affects contact creation, it seems like it creates multiple ones: the flat, 3DEC-created model has 1 contact and 934 subcontacts while the bended, Salome-imported one has 14219 contacts and 5940 subcontacts. Note that a third model with a flat contact meshed in Salome then imported into 3DEC gives equivalent results than original flat-3DEC geometry but has multiple contacts and subcontacts as well).

Thanks a lot for your explanations!

Cheers

Theophile

2

When you use “merge”, every triangular face on the outside of a block is a separate object with its own nodes and its own contact. This allows for complicated (concave) block faces (usually these are coming from Griddle or similar). However, this also means that the joint plotting may look a bit weird as you have seen. Because each triangle has its own set of nodes, it is possible for there to be two different values at the same node location. So the contours can appear discontinuous. I guess we should do some sort of averaging to smooth out the plots - I need to think about it.

3

@jhazzard, thanks a lot for the clarification!

When averaging the results on subcontacts, maybe it would also be of importance to make sure the contact laws are preserved (consistency between averaged u_n and F_n)?
I think this would keep physical consistency, but I am not 100% sure this is relevant.

Another possibility might be contouring according to triangle centers where normal values would be the average of the values on triangles subcontacts?

Cheers

Theophile