Multi Body Impact

[Phreakish]

CosmosWorks Advanced Pro 2008.

the ultimate goal of this is to study the interactions of a two body sliding assembly. one part is fixed, the other slides in 'guide rails' and is sent into action by a shock load (think automatic rifle).

In preparation for this, I've been trying to familiarize with Cosmos capabilities, and before purchase of the advanced pro, I received a demo vid of two parts interacting. The 'moving' part was given an initial velocity, and the 'stationary' part was fixed, and the animation shows the two parts colliding and the stresses resulting in each part as a result.

I'm sure its something simple, or a misunderstanding of terms. I have two blocks (two instances of the same part) with a material applied, one fixed, the other fixed in all rotations but able to translate (at its initial velocity) toward the fixed block. If I'm able to get results, the animation shows the blocks stuck together at the impact face (there are sufficient steps to show translation before imapct) with the stresses increasing, but both parts are stationary. What must I do to 'see' the translation before impact and the resulting velocity after impact?

 

[Phreakish]

I setup another simple excersize to try and trouble shoot. 4 hrs to crunch this thing.

Two parts, one slides in the other, the point at which they should impact, they dont. Rather, the moving part that should impact the first, travels FAR clear of the stationary part, and THEN out in the middle of space it sees a dynamic impact and rebound. Why is the contact 'found' way out here? why is it not found when the two parts actually interface? This is odd, and frustrating.

Has anyone else experienced this?

 

[cbrn]

Hi,
my explanation refers to how CW worked back in 2004 / 2005 versions so it may have improved in the meantime.
Back in these days, CW's contacts were "linearized", i.e. the surface contact properties (relative stiffnesses, pinball etc...) were evaluated based on the initial state configuration, then never updated during the analysis. The impact of a body against another is, instead, a highly non-linear phenomenon. If CW is still set up to a "linearized" contact, then it's no surprise that the initial evaluation of the relative stiffness is totally wrong, since in the initial state the two bodies are "far" one with respect to the other. This explanation is rather simplistic, the math underlying the contact algorithms is quite complicated, however this should give you an idea of why the simulation isn't performing as expected by human - and physical - logic. Back in 2005, your simulation would have been almost impossible. Perhaps with 2007 version there are some more controls for the contacts: look especially for anything which could sound like "predict for impact", "update stiffness at every timestep", "auto-adjust stiffness" and so on.
Regards

 

[Phreakish]

Thank you for the explanation. I'm using the non-linear, dynamic, section of 2008 - I've been assured that its fully capable of this type of simulation, but I'm not completely convinced as of yet.

I've gotten the two bodies to impact, but here's another oddity. Depending on the distance between bodies prior to impact, and on the time step used, I seem to get random changes in the end velocity, and the end velocity is typically HIGHER than the velocity prior to impact; very odd.

 

[DLTguy]

Hi
For such problems ABAQUS or LS-DYNA is appropriate. If you want to get results (numbers) that you can trust and usable, then COSMOSWorks is not appropriate for impact. If you want to see just a fancy video with colors then perhaps COSMOSWorks is okay.

Regards,
DLT

 

[appliedocean]

Since you have Cosmos Pro,
Try a "Drop Test Study" using only your sliding part. It sounds like this will get you half the way there because you are only using the "moving part" for the study. However it may provide you with enough data to make design decisions.

in Cosmos search for "drop test"
Good Luck-

 

[bubbazee]

Can you use COSMOS Motion to get the impact forces then transfer this into COSMOSWorks? I am not exactly sure what you are looking for at impact, so forgive me if this sounds like I am simplifying the study.

Thanks,
Ryan C

 

[Phreakish]

Thanks for the tips folks, here's what I did:

Cosmos works, and the data fits with initial hand-calcs (within reason). But its slow to process nonlinear, dynamic models, and my results should still err on the conservative side.

I modelled the two parts at a very small distance apart (just prior to impact). I couldn't use as fine a mesh as i would have liked, but am well within reason.

Instead of looking at forces and stresses caused by the impact, I instead gathered the time it took for the impact to occur, used that to extrapolate the acceleration at impact, and then re-ran the component with that acceleration applied, appropriately constrained, and an adaptive mesh. Several averages were taken for the accelerations, since it was non-linear over the event, and the highest average chosen for the final modelling. Since only one of the two components was of concern, that was the one investigated.

As far as cosmos motion, I've found it useless (for my purposes at least) for anything but mechanical simulations to see how parts move (if cosmos says they'll move, they likely will, since it seems to detect a crash anytime two parts seem to interact... frustrating).

I did look at the resultant force at impact, averaged over a generous time interval and compared that to the initial theoretical forces, and they were within 13%, which I deem reasonable.

The best news is that the parts survive for life cycle expected.