symmetry boundary condition

suktan

Can I use frictionless, elastic or displacement constraint interchangeably to infer symmetry boundary?

Victor

In principle, yes - they have the same physical meaning. Here are some differences:

• Elastic support requires you to tune the stiffness parameter to be high enough that it's almost rigid and low enough that the solver doesn't fail.
  
• Elastic support doesn't currently work with the CalculiX solver.
  
• Frictionless support follows the orientation of the element surfaces so you have to make sure they're all in the symmetry plane. Usually they will be but a particular case where it can go wrong is the edge of a curved cylindrical shell which will usually be slightly tilted from the symmetry plane.
  
• Displacement and frictionless support are both incompatible with bonded contact on the same nodes.
  

Frictionless support is usually the easiest and most versatile.

suktan

Thank you, Victor.

suktan

I need to use elastic support for the presence of bonded contact. Is it a trial and error step to adjust the stiffness parameter? I can set it as high as possible as long as the solver converges. Correct?

Victor

Yes, trial and error. A reasonable starting value is young's modulus divided by a characteristic thickness, such as the thickness of an element. Or you can imagine there's a thin layer of extra material between the two parts, and use the thickness of that.

I might have been wrong before about too high a stiffness. When it's something like 10 orders of magnitude above that starting value, it seems to cause the boundary to become overstiff in other directions so the solution would be wrong.

Rusticus

Victor -

Which "characteristic thickness" is appropriate? That of an element, the thickness of the part (in one direction or another), or something else?

TIA

Rusticus

Victor

I suppose element thickness in the direction normal to the surface but really it should be the thickness of an extra layer of imaginary non-space-occupying elastic material you can put between the surfaces without worrying that it'll deform significantly.

Don't spend time measuring it though since it only needs to be within an order of magnitude or two and you'll likely have to adjust it anyway.

Rusticus

Victor -

Thanks for further elucidating.

I may need to use the elastic support constraint, since I am attempting to model 1/4 of a revolved pair of solids which have an interference fit on the ID/OD. Thus I would have nodes with both the symmetry condition(s) and bonded contact applied. However, I am now off to experiment with thermal loading to achieve the interference pre-load, rather than (exclusively) using the bonded contact.

Thanks again.

Rusticus