OK. A lot of views but not too many replies, so maybe the best way to address this issue is that I simply put small flat cuts on the interfacing parts, to simulate the initial contact region, then mate those flats to avoid the issue of interference modeling. Is this the best approach for a problem like this? Is this how one would typically attack modeling the dynamic performance of a contact point in an assembly?
My initial concern is deciding what that contact area should be, but that might be only a secondary problem, and I could vary this parameter and see if it has any strong influence on the final result. Just a thought for maybe some feedback.
Maybe the primary problem with what you are trying to find is that No Penetration and Shrink Fit contacts are a kind of nonlinearity where the Frequency solver only solves for linear solutions. This is why you don't find No Penetration/Shrink Fit as options for a frequency study. What you can do though is run an upper and lower bound analysis. In one setup you run an analysis where you have bonded the sphere to the cone. In the other setup you allow penetration between the bodies. In this way you are running the two potential outcomes of your real life setup where they either 1) are stuck together, or 2) broken apart. In real life the natural frequencies will be somewhere between these two setups. More realistically though, if you are making an interference fit, the bonded analysis is going to be closer to the real life conditions.
Also, cutting flat faces for the contact areas as you describe in the second post would be a good setup for initial contact while negating the interference and is a good idea to implement. I hope some of this helps.
Thank Ryan for the input. Yes I did run an upper and lower bound model using a range of contact area with the "cutting flat faces" approach. It seems to be bounding the problem properly now and I can test against actual measurements to narrow the model in. I think this is what may work best. I just have to set my Mesh Controls dense in the regions of contact, but that's not a problem.
I was kind of hoping to have a contact region that would evaluate contact stiffness in the model as well as be free to rotate with friction control to model other details in the model, but my rotations are small and maybe not large enough to be concerned with. Thanks again.
If you have test data to compare against it may also be realistic to put a stiff spring between the ball and cone and calibrate in the stiffness of the spring to approximate an imperfect bonded contact.
Just another thought.