Contact Interface connections

[ptc-4877204]

I am having trouble getting contact interfaces to function as expected in a standard/static run. I am evaluating the deformation of a rubber mount under axial and radial load. I have theorized all the potential contacting surfaces and manually defined those interfaces, but the adjoining components continue to visibly pass through one another, upon examining the displacement results.

How can this be eliminated? Also, why is the gap element not available in Wildfire 4.0?

I appreciate any assistance. Thanks.

[integrisbrad]

You can verify whether or not contacts are actually active by looking at a quantity like contact pressure. Sometimes visualizing exaggerated displacements can be misleading when it comes to contact.

I definitely remember instances where the contact was not detected if I had contact surfaces that were a significant distance apart initially...and even occasionally when they weren't apart . In that case, some different attacks to the contact definitions or small shifts in the geom would often resolve the problem.

Mechanica doesn't have a whole lot of controls/knobs to turn when it comes to contact, and that was something that we requested in speaking with their development team in the past.

The following suggestion may or may not work. It does help in other codes, but maybe for different reasons. If you activate time stepping instead of applying the load all in one step, it may recognize the contact surfaces that have some more significant initial gap since the intermediate steps will get the surfaces closer together. Or, Mechanica might just make that active/inactive determination up front and the suggestion might not help at all.

As far as gap elements, I can't answer. I think in later versions (beyond WF4) you can tune nonlinear springs to function somewhat like gap elements, but I never really investigated thoroughly. Behind the scenes though, I believe contact may be creating something similar to a coating of Nastran type gap elements in Mechanica.

[ptc-4877204]

Thanks for the reply. I have tried load-stepping and haven't really noticed much difference, other than the long solve times. But that would make sense, as a possible option.

Also, can you possibly elaborate more on what variations you may have tried in the past on the contact definitions to get them to function more predictably? I have tried adding infinite friction, for example, and sometimes that helps and sometimes not.

There must be a way to get some indication that your job is not going to solve properly, before the full run time elapses (6 hrs., for example). Mechanica seems to predict possible/impossible contact before the job begins, so there must be some check in place to evaluate your contacts at the start(?)

Also, how are these elements improved in WF5?

[integrisbrad]

Sometimes just deleting and redefining a contact works, small shifts in geometry position, slicing up conact surfaces with volume regions, deleting and redefining existing volume regions, order of selection...all some of the stuff I've had success with in the past.

And no, Mechanica didn't seem to give much of an idea of ultimate success before the analysis ran. It would often tell you that contacts weren't active in the up-front messages, and then they would still show up as being active in the results.

To cut down on run times, I made liberal use of weak ground springs to stabilize the contacting parts. That was the only thing I found that really helped (other than general model robustness...other BCs, etc). For some models when I started out using contact, it seemed that creating a small initial penetration was favorable, but that turned out to be false as a general statement.

I believe I recognize your name from my former company, Greg. If so, some of these suggestions might be out on a wiki site in the FAQs I left behind.

[ptc-4877204]

Brad--Thanks again for the reply. Your comments and suggestions are helpful. I do think, in general, that perhaps a different tool would be in order, on similar future nonlinear contact analyses. NX Nastran, for example, may be a better choice, would you agree?

I recognized your name as well, on previous posts, and was hoping for a reply. I have moved on as well, so I'm afraid I would no longer have access to those wiki FAQ's. Looks like your new company has a full and impressive portfolio of services. Congrats

[integrisbrad]

Sorry for the late reply, but yes, I agree that there are better tools for contact analyses. For those who don't have those tools available, Mechanica can stand in as long as the limitations are acknowledged.