Convergence of shell elements

[jacopo.gili]

Hi all,

A fast question, I have a model where both solid and shell elements are used. It is a model of a Pressure Vessel. The Vessel itself was meshed with shell elements and the flanges and Nozzles with Solid elements. I am excluding from convergence the nozzles since there is some singularity there due to contacts between the shell and solid elements (i suppose, BUT THIS IS NOT THE POINT). Nevertheless if i only have shell elements selected for convergence (having excluded all the parts meshed by solids) no H refinement is applied.

Seems that H refinements are not applied to the shell structure when also the solid elements are present (even if excluded from convergence). If i exclude them from simulation than the refinement works fine, but i need them since they constitute a reinforcement of my holes.

Is this some sort of bug?

Best,

Jacopo.

[raviburla]

Hi Jacopo,

H-refinements on shell structures are supported by design. From what I understand from your post, if the initial run did not have convergence solution, then atleast one h-refinement should be run.

Can you check few things:

1. What convergence rate does convergence plot show. Is this consistent with the convergence settings.

2. Change the convergence settings (reduce target convergence percent) and see if that helps.

If the problem still persists, can you post a simplified model to demonstrate the issue? 

Thanks,

Ravi Burla

[jacopo.gili]

Hi again Ravi!

The convergence rate on the convergence plot is 50% 1 step, the classic one which is always displayed when no h refinment are applied. However i am applying 4H refinment with a 10% Stop Criteria.

Figure 901 shows you the model. The green parts are those that are meshed by solid elements but are excluded from convergence. The rest (which is a thin body recognized by inventor) is meshed with shells. I would expect the convergence to be calculated for the shells siche all the other parts were excluded from convergence.

Figure 902 shows the convergence plot: 50% convergence rate and 1 step (P refinment i understand); no H refinment are applied! Even if i asked for 4 and the convergence rate is 5 times higher than the one i require.

At this point i did 2 things:

1) excluded the parts meshed with solid elements from the simulation (not from convergence): in this way i have only shell elements in my assembly since the parts that would me meshed with solid elements are not present. Here the convergence starts. (i did not really see if it was achieved or not but just concluded that it starts at least).

2) I made a new simulation where i meshed everything with solid elements. And there it happens a strange thing! the convergence works and the convergence results (the maximum VM stress for example) are numerically in good agreement with the one obtained by the first shell model.

This makes me think that the first model was in fact converged but without the solid mehed model i could not be possibly sure of my result (due to the first convergence plot being quite nonsense)

Best Regards and thanks in advance for your help,

Jacopo.

[raviburla · Accepted solution]

Hi Jacopo,

Thanks for the details. I was also able to reproduce the issue on a simple model as well -  Excluding either solid or shell from convergence for a mixed model does not lead to h-refiements.

As a workaround, you may want to apply local mesh refinements in the regions of interest, or include all the bodies in convergence.

We will log this issue internally in our defect tracking system.

Thanks,

Ravi Burla

[jacopo.gili]

Hi Ravi,

I am happy to hear that my doubt leaded to the aknoledgmewnt of an actual bug. Now it will be fixed and not a problem anymore!

Regarding your workaround suggestion, I actually did that. It worked just fine!!

Thanks!