[Tensian]

Dear foamers,

I was wondering about the options for modelling contacts in foam extend and also in solids4foam. I found the FVM approach for this very interesting...

For example, which condition will be the "equivalent" to a bonded contact (like for example in ANSYS Mechanical)?

Thanks a lot in advance

[iBatistic]

Currently, solids4foam treats contact with a penalty procedure, which is fundamentally similar to the node-to-segment approach from fem. The procedure is suitable for small and large deformations and for rigid-deformable contact.

In bonded contact, bodies are glued and there is no sliding and separation. I think that this behavior can be easily achieved using GGI or AMI, so there is no separate implementation for bonded contact.

[CRI_CFD]

Originally Posted by iBatistic

Currently, solids4foam treats contact with a penalty procedure, which is fundamentally similar to the node-to-segment approach from fem. The procedure is suitable for small and large deformations and for rigid-deformable contact.

In bonded contact, bodies are glued and there is no sliding and separation. I think that this behavior can be easily achieved using GGI or AMI, so there is no separate implementation for bonded contact.

Thanks for the suggestion

[nos_cfd]

Hi!
I have also been involved with penalty vs bonded contacts recently. I am working on a similar problem and have as reference results from running simulations with bonded contact. Have you succeeded mimicking the bonded condition with foamextend / solids4foam ?

@iBatistic I have been reading your recent work : A finite volume penalty based segment-to-segment method for frictional contact problems, Ivan Batistić a, Philip Cardiff b, Željko Tuković.

It seems that this approach is better at controlling penetration. Does the segment-to-segment approach mean to replace the GGI ? Is this the current underlying mechanism for solving the contact penalty in solids4foam v2.0 ?

I read in related threads that to handle numerical instability, low relaxationFactors (in the firction & normal Models) should be used. But they allow more penetration so it diverts from the bonded condition. I was wondering if with the segment-to-segment model is more stable and therefore, we don't need to increase the penetration gap.

Thanks for sharing !

[iBatistic]

Originally Posted by nos_cfd

It seems that this approach is better at controlling penetration. Does the segment-to-segment approach mean to replace the GGI ? Is this the current underlying mechanism for solving the contact penalty in solids4foam v2.0 ?

The segment approach is more related to penetration calculation. Penetration is integrated for each face in contact instead of taking values at face vertices. By doing integration, slave-to-master interpolation is avoided, and some problems related to vertex calculation of penetration are solved.
It is currently unavailable in solids4Foam, but I guess that will be anytime soon.

Originally Posted by nos_cfd

I read in related threads that to handle numerical instability, low relaxationFactors (in the firction & normal Models) should be used. But they allow more penetration so it diverts from the bonded condition. I was wondering if with the segment-to-segment model is more stable and therefore, we don't need to increase the penetration gap.

Low under-relaxation factors are needed to allow for the stable augmentation of contact traction in between iterations. Their value only affects convergence speed, i.e. the required number of iterations to reach the desired residual. The accuracy is controlled by the penalty factor, which is a user-defined violation of contact conditions. A bigger penalty factor improves accuracy but reduces robustness and vice versa.

If you want to model bonded contact (no sliding and no separation) why not use the multi-material interface treatment described here (it is implemented in S4F):
Finite-volume stress analysis in multi-material linear elastic body (Tukovic, Ivankovic and Karac)

[nos_cfd]

Excellent. I hope the segment-to-segment gets integrated soon and used. Sounds very interesting. Thank you very much for making it available !!

In my case, I treat the geometry by regions and mesh each body independently so contact regions are not conforming. I thought the only way of imposing the contacts was through the normal and friction models in the D / DD files using the relaxation and penalty factors for penetration. Thank you for the paper that you suggest, I will have a look!

...Although I am starting to think that bonded with non-conforming contacts doesn't seem feasible...

[iBatistic]

In bonded contact, the contact area is known in advance, which means that the contact loses its main source of nonlinearities (this fact is useful for implementation). Currently, you are using a boundary condition that is developed for the unknown and evolving contact area.

Solids4Foam can use cell zones to treat mesh zones as different materials, but I think that the same procedure with non-conforming separate meshes is not available. Probably should be best to implement a separate boundary condition to treat bonded contact.

[nos_cfd]

yes, that's what I figured... I am currently using cfMesh to generate the meshes and I refine through regions (patches) which have identical .stl files except for the triangles orientation. The resulting meshes in the contacts have almost the same facets. From that, I think it is possible to have an almost conforming contact. I will try implementing the bonded contact.


Thanks for the ideas!!!!!

[Youngxl]

Originally Posted by iBatistic

In bonded contact, the contact area is known in advance, which means that the contact loses its main source of nonlinearities (this fact is useful for implementation). Currently, you are using a boundary condition that is developed for the unknown and evolving contact area.

Solids4Foam can use cell zones to treat mesh zones as different materials, but I think that the same procedure with non-conforming separate meshes is not available. Probably should be best to implement a separate boundary condition to treat bonded contact.

Hi Ivan,
Recently, when I used the contact model in s4f to calculate my case, From function void newGGIInterpolation<MasterPatch, SlavePatch>::calcAddressing() in file LnIclude/newGGIInterpolationWeights.C at line 293 The master projected polygon was CM instead of CCW. This is strange... appeared. I am curious why this problem occurs and what can be done to solve it? My case is about the extrusion of a ball on a V-shaped surface in plane strain state.
Hope you could give me some advice.
Best,
Young

[iBatistic]

Hi Youngxl, that is strange, I never got that warning, it occurs because the boundary face points are oriented in the wrong direction. Can you share the case?