Convergence Question screw in cylinder

General Questions
1

Hello everybody,

I have a questiion about convergence: I have a simple model of a tiny screw in a cylinder and I’m trying to show the max. stress and displ. differences (only on the cylinder) when the moment applied to the screw is misaligned. For now i am tilting the moment from 0 to 20 degree in 5 degree increments. The screw is tied to the cylinder by tie constraints (only the thread). It is very simplistic but i only want to show that a misalignment has any effect. Since it is my first FE study im a liitle bit lost on the convergence.

grafik
grafik706×370 74.8 KB

What I already tried:
I repeated the same FE simulations with different mesh paramaters to see the differences the mesh setup has on my results. I started to refine the mesh by incrising the elements per edge and curvature and decreasing the min. element size (to get a finer mesh around the screw-cylinder interface).
Then I calculated the error between each mesh in relation to the finest mesh and ploted it against the different mesh setups (for the DOF i just multiplied the nodes from the simulations output by the DOF of each node).

The max. displacement is converging quite nicely but the max. von mises stresses aren’t at all.

Now my questions:

  1. Is there a better way to show convergence than what I did?
  2. Why do I sometimes get different node numbers with the same mesh parameters?
  3. I cant increase the detail on the mesh anymore because calculix gives me an error (I think it is to fine now)?
    grafik
    grafik1824×578 62.3 KB
2

Check this: FEM Geometry Preparation and Meshing - FreeCAD Documentation

You might have encountered a stress singularity. You would have to check where the peaks occur, try refining locally there and consider the workarounds mentioned on the aforementioned page if it stil doesn’t work.

3

If I were doing this, I would start by simplifying as much as possible. If the model is axisymmetric, it would be easier to partition it so you can easily mesh it with linear quads. In my experience, tets are not very good at showing convergence with contact. So, break it down into very simplistic partitions, then sweep your mesh and test refining it.

4

studying convergence of mesh in linear elastic materials can be hard to achieves due to singularity. Switching to non-linear material and plasticity goes better since it will redistribute as in reality. My previous test in comparison of seam weld model in part connection using tie constraint, tied contact and continuous mesh. Linear elastic analysis produced many stress spots for tie constraint models, but result much better in nonlinear analysis with plasticity, no have large discrepancy for all of them in equivalent plastic strain.