Problems with mixed mesh: plates/beams

[pandaleone]

Hi all
I'm having a problem with a mixed mesh composed by beams and plates for a static analysis. The system is well constrained but I receive errors typical of meshes with mechanisms, i.e. high pivot ratios.

The structure has a (qualitative) shape:

------------------------------------------------
| | | | | | | | | | | | | | | |
------------------------------------------------

where vertical signs are beams and horizontal signs are plates (obviously the structure is 3D). I have tried to turn the parameter K6ROT on and off, and I have varied it 1.E-5, 1, 1.E+5 but nothing changes. AUTOSPC is turned on.
The plaets are parabolic CTRIA and using CTRIAR do not change the result.

I have tried modeling a small portion and have tried a modal analyis and I can see beams rotating "freely", although the K6Rot should avoid that.

When I tried to give
PARAM,BAILOUT,-1
the analysis run without further problems and the result was reasonable. I don't want to use rigid elements around the beams beacuse the beam pattern is very complex and too many meshing efforts would be needed. Unless there is no way to adjust this analysis.

Hint: There is a part of the model were other plates of larger dimensions are present; may the high pivot ratio be due to big stiffness differences between these two regions?

 

[JakeLiv]

the beams are 2d and the plates are 3d? or the other way around?

in both cases, you need an overlap of shell-on-solid elements to avoind a mechanism effect.

 

[pandaleone]

Hi

the structure is 3D, I only sketched an idea here. Imagine two disks meshed with plates which are kept together by those beams. The only dof which is free is RZ for plates, that is RX for beams; for this reason I expected that K6ROT would be enough. Beam torsion is NOT a load path, so I'm not scared of using K6ROT, but I would like to know what else I can't see here is causing the errors.

 

[GBor]

Just as a check, use a low-stiffness beam in the plane of the plates connecting the end of each beam to the plate at a node adjacent to where the main beam intersects the plate. Generally, this is called a "spider" and is used to handle the DOF issues.

I understand k6rot was perhaps designed to handle this, but this would at least tell you if it is working or not.

 

[ThomasH]

Hi

Just as a simple check, run a eigenfrequency solution for the structure. If you get results with modes near 0.0 Hz you have free-free conditions but you will get a plot of what is it si that is free.

The limitation to this is that if you have many modes that are zero the solver probably won't give any results at all.

But it might be worth a try.

Good Luck

Thomas

 

[pandaleone]

Thank you for your replies.

@ GBor: Your solution is the best for sure; but I am operating on a complex unstructured meh, where the beams (more than 1000) are positioned on a complex patter: to add a spider I should remodel the whole structure to position support nodes in a proper way; or I should add a thousand of spyders by hand. I wanted to reach a solution which did not require heavy manipulation of the model.

@ThomasH: I've already run a modale test on a small portion of the model and the problem seems to be exactly what I suspected: I get free (or almost free = 10^-8 and 10^-5 Hz) rotations on the two beams I've kept.