Contact issue work around

Fatmac

For some time I have struggled to get many of my contact models to run where there is an initial gap present. I have played around with gap materials, different methods or loading the model (force and position control), elastic boundary constraints, moving parts till they touch at the start of the run, and some other wild ideas. Although I generally get the analysis to work in the end, I'm often not altogether satisfied with the approach/solution. Using the adjust card will normally resolve the issue but frequently it is not what I'm after as the gap is part of the geometry I'm trying to model. Any help or tips would be much appreciated. Just recently I've stumbled across a potential work around. I'm working on a 1/2 model of a car running on a track connect by a nylon rod. The rod is under sized and as normal I was having trouble getting the analysis to run. I tried some of my normal tricks but nothing did quite what I wanted. I finally tried attaching a very light spring (25N/m) between two parts in a model where the loads are in the order of 500 kg). To my surprise the model ran as expected! Varying the spring stiffness does change the speed of convergence a little and in this case below a stiffness of about 5N/m convergence is very slow. I did try adding a similar spring at the other end of the rod to see if it would help mode-again to my surprise it stopped it converging! I'm wondering if this is a one off oddity or could be used in other similar situations -I will experiment more but in the mean time would be interested if anyone has had and similar experiences?

Victor

It looks like the common problem of rigid body motion of unconstrained parts. To reliably converge, the rod should be constrained in all 6 degrees of freedom. Elastic support with both normal and tangential stiffness (tangential stiffness is zero here) can do that, or 6 springs. One or two springs might not always work because that's not sufficient to constrain it fully.

Also, increase the contact stiffnesses by a factor of 100 or so because penetration is very high.

I got convergence with:
- No spring
- Elastic support with 100000 Pa/m normal and tangential stiffness per unit area.
- 55500 GPa/m stiffness per unit area in the contacts.

You would still need to make sure the elastic support isn't restraining it too much.

Fatmac

Thank u again Victor. That all makes sense. For most of my applications I think I like the 6 weak springs connected to a part that will be sort of moving with the free body rather than being connected to earth which is what I understand the elastic support is affectively doing. Its also easier for me to see how big an affect each spring is having on the result be simply looking at its deflection. I've been playing around with the simple model I posted above and with 6 springs the analysis converges even with zero stiffness 0 N/m! seems a little odd.

Victor

Yea that's right elastic support connects to ground so springs may be more realistic if the parts move together. Ideally, you'd use such low stiffness that it doesn't matter but that might not always be possible.

It does sound strange that 0-stiffness springs affect it, but convergence for completely free rigid body motion sometimes does seem to be a matter of luck and not something to rely on.