COSMOSMotion does not support deformable parts, and in general the 3-D contact assumes that the penetration that occurs will be small. That being said, there are a few approaches you can take to model in essense a sticky collision.
If you are still using SW2007, then the impact force is a good way to investigate collision settings to get a the sort of behavior you want. The impact force works similar to 3D contact, it is just a matter of specifying a minimum distance between 2 points that the force starts to act at and then the nonlinear spring/damper characteristics of the foce.
The main factors to vary are stiffness, exponent, and damping. The exponent term will obviously adjust how quicky the stiffness increases with penetration, and the damping will determine the energy loss during the collision as well as the velocity dependancy.
The main trick for the impact force is to use a reference point well behind the depth the part will penetrate in so that when measuring the distance, it is always going to be in the right direction.
You can do the same with 3-D contact, but it is a little trickier. With 3-D contact, the interference is computed between to bodies, and the force direciton is determined by projecting the center of the interfernece to the nearest surface. So if the parts penetrate a little, it works great. If they penetrate alot, it may find the a back surface closer to the center of the interference volume and apply forces to push it out in that direction. If you create a really deep solid such that for the desired penetration, the nearest surface to the interference is the once you want the part to go back out of, then you can do it that way.
If you really want to make sure a part remains in contact once it gets within a certain distance of another part, you can use an approach similar to modeling magnetic fields or air vacuum. This is where you write an action reaction force expression that measures the disstance between the parts and once the distance is less than some criterion, it applies an attractive force to keep them together. It is best used in conjunction with contact or an impact force to clamp in place.
When dealing with contact and high speed collision, the solver settings can improved the accuracy. Make a maximum time step allowed smaller by an order of magnitude or even two (it needs to be small enough to capture the collision with sufficient resolution).
I alway recommend trying this out on a simple test model to get a better understanding of the parameters and how you can get the to model different material responses.
I hope this gives you a couple of ideas.
Do you have an example of the motion analysis you described? I'm trying to do an impacted collision but parts keep flying through each other. I have a known velocity and force but the objects go right through eachother because I don't know how to turn off the velocity after impact.