N-S Coupling vs Refinement

cwharpe

Been investigating the use of RBE's in Mecway and other softwares. I may have stumbled across a potential pitfall when refining a model with Node-Surface Couplings (picture below). Pays to check displacements after the solve.

The issue seems to be when the coupled surface (e.g. hole perimeter) has a varying element resolution across the thickness -- like what you get when spot-refining a solid shape. Not sure of the causal mechanism that is skewing the results. The way I managed it was sub-optimal because I had to decide the local refinement before I extruded into a solid.

Comments? Anyone else experience this?

Victor

I'm not seeing that asymmetry. Could you post the model where it happens?

The fact that it depends on the order of refine and extrude suggests something's wrong with the mesh.

Are you sure the reference node is still at the center? If you create it with Insert node between on the non-uniform mesh, it won't be at the center.

cwharpe

File w/ Both cases together attached. Centers checked and projected cylinder surface checked. Same asymmetry present.

Is the rigid coupling trying to act like an RBE3 (distributed), with the weighted average of extra lever arms over the Refine spot thickness becoming a factor? Just guessing here.

Victor

I can reproduce it now. It seems to be because of the pyramid elements at the edges of the refined region. I'll investigate further but it might turn out that the solution ends up being "don't use pyramids with node-surface coupling"

Mecway just exports the surface to CCX as nodes without weights, so there's no clever accounting for the areas of the faces. But pyramids are really collapsed hexes in CCX so maybe those duplicate nodes somehow mess up the rigid body MPCs.

Thanks for discovering this, cwharpe. It's an insidious bug.

I also tried *COUPLING with *KINEMATIC instead of *RIGID BODY but it's no better.

cwharpe

Yes, I see that now. You seem to be on the right track.

Thanks.

JohnM

You know the joke: The patient says, "Doctor, it hurts when I do this." Doctor says "don't do that". I have learned to avoid refining hex models that require making pyramids. I have seen things that make me suspicious of their behavior. I've not seen something as obvious as this, but I have seen models that will not converge as well so we "don't do that".