Discussion Forum

Olivier-

Try using the Continuation solver and ramping the prescribed displacement from 0 to -0.6 over a range of values. This will help the solver. I tried it for your problem and was able to engage the contact. Although, the final parameter step was having some difficulty the contact still was engaging.

See attached *.mph file.

--
Best regards,
Josh Thomas
AltaSim Technologies

Olivier-

Try using the Continuation solver and ramping the prescribed displacement from 0 to -0.6 over a range of values. This will help the solver. I tried it for your problem and was able to engage the contact. Although, the final parameter step was having some difficulty the contact still was engaging.

See attached *.mph file.

--
Best regards,
Josh Thomas
AltaSim Technologies

It works perfectly ! Thank you. I did not know this tool and it seems useful.

For the last parameter ramp, Why it is difficult to converge? In this case, how to speed up its convergence, to subdivide the range again or use other method? It seems like tricky at this point.

I had a similar issue and found I got superior convergence by not using the built-in contact model, which is quite general, but rather using a coordinate-dependent boundary force. For example, if the boundary is at x = x0, then the force could have the form:

(x > x0) * (x0 - x) * k

where k is an effective spring constant, and it is assumed the contact force is in the -x direction when the materials overlap.

A disadvantage of this is the target is modeled ideally, and it won't deflect, thermally expand, etc. However for my application this was fine. Additionally it allowed me to readily implement a inelasticity model for the impact, reducing bouncing.

Hi, Daniel,

Thanks for your share. It's novel to discard the regular contact node and substitute it by boundary load. For the specific formula to impose, how can I know it and can you give me some references. In addition, for the contact of sphere or circle, the formula you mentioned is applicable to all cases or not?

Thanks,

Zeng

Hi,

The formulation suggested by Daniel is essentially the same as the built-in Penalty formulation. In COMSOL Multiphysics, a force which depends on the displacement automatically ends up in the stiffness matrix.

The built-in version does however have a number of advantages:
- Works for flexible and moving bodies
- Can have friction
- Predefined postprocessing variables, like contact pressure and gap.

This said, there are some interesting applications of Daniel's suggestion:
- I sometimes use it when there is contact at a symmetry plane. You do not have to create an extra fictitious body there as when using a contact pair.
- As Daniel says, you can more directly tweak the reaction force to make it depend on for example time or velocity. It would be interesting to see some more details about the "inelasticity model for the impact".

Regards,
Henrik

Henrik:

I also implemented friction and inelasticity with my simple approach. I was tempted to implement adhesion as well, but I lacked sufficient confidence in the analytic form for this. It would have worked with transient simulation but not with the continuation method. Comsol's built-in adhesion model, per 5.2a, also provides hysteresis with continuation.

But the motivation in my case was convergence. I just couldn't get the built-in model to reliably converge, even following the recommendations on mesh. I'm sure there was a solution, but given limited time it was easier to just use an explicit coordinate-dependent boundary force.

Of course, this requires the luxury of a relatively simple contact, like a perfect planar target, or else you're in the regime of analyzing surface normals, and I'm sure the internal model can handle that complexity more efficiently than some home-brewed equation.

I have a similar issue. I'm doing a stationary simulation of a bracket using Solid Mechanics physics. The intention is to know where there are the highest stress and compare with data obtained from an experiment.
But the simulation is not converging. I've tried several different ways and it does not converge. And this is because the contact restriction is not working. The contact faces pass through the others. I've even tried using Multibody Dynamics and also the contact does not work.
Does anyone have any idea how to solve this? The simulation files are attached.
Sorry for my English.

Thank you!