Transforming thin sections to analyze with solid meshes

[imagitec]

Time didn't permit me to solve my last problem of extracting reaction forces from a shell model to use as an input in the solid analysis of the mating part. I will revisit it, so I know how the next time. I'm still accepting advice, however.

While discussing a similar thin section/fat section problem with another engineer, I thought of a potential workaround for CosmosWorks' inability to mix solid and shell models. I haven't thought it through, but I'd be interested to hear your thoughts, especially if you've tried something similar.

One closed form solution for analyzing a composite beam (shaft) is to transform the area (polar moment of inertia) of the stiffer member(s) by a factor proportional to the ratio of the stiffness of the stiffer member to the summation of all the stiffnesses (I'm not certain if the preceding is correct; I don't have a reference in front of me). When analyzing a beam, such as a wooden beam with metal plates on the top and bottom, it's important to transform the area of the plates by increasing the width and not the height, because the moment of inertia is proportional to the cube of the height.

I was wondering if a similar approach, modeling a thin section as a thicker one with a lower modulus of elasticity, might allow it to be modeled with a solid mesh. As with the composite beam above, the relationship won't be linear in many cases, but it might be possible to develop valid relationships.

Maybe CosmosWorks will support transitional meshing in the next release.

Thanks,
Rob


Rob Campbell

http://www.livejournal.com/users/robcampbell

 

[corus]

The bending stress in a shell is inversely proportional to the thickness squared whereas the direct stress is inversely proportional to the thickness. Increasing the thickness and modifying E in the shell won't give the correct solution.

It would be difficult to tie a shell to a solid part as the rotations of the shell could not be linked to the solid without using 2 layers of nodes at least. in other words, tying a single node of the shell to a node on the solid would be incorrect as the rotational freedom would be free.

If I wasn't interested in the stresses at the junction of shell to solid then I'd perhaps bury the shell into the solid so linking all freedoms. If that option isn't available then I'd think about just modelling it all as solid and accept the small elements at the junction.

corus

 

[imagitec]

Thanks for the reply.

The problem I'm facing in CosmosWorks is that shell and solid meshes can't be mixed, and the meshing or subsequent analysis fails when I model the thin section (a bellows welded to the solid part) as a solid. The size of the elements in the bellows determines the size of the elements in the whole model, so the resource requirement is large.

The best solution is to get better software. But now I'm interested in the theoretical exercise (which may lead no where).

Thanks again,

Rob Campbell

http://www.livejournal.com/users/robcampbell

 

[corus]

If it's at all possible you could replace the bellows by spring elements with the correct stiffnesses in the required directions, that is presuming you can attach beams to solids. This wouldn't represent local effects properly but you might get the right global response to the larger solid part.

corus