Thin shell assembly meshing difficulties

[CalculatorDave]

Howdy all,

I have had consistent difficult in achieving proper meshing of thin shell assemblies in COSMOS. When I run the model, I get a message saying that there are "insufficient constraints", meaning that the parts of the assembly are not bonded together, at least as COSMOS is concerned. Even though I create lots of split lines (to try and force a continuous mesh at the joints of parts), and even though I establish bonded contact conditions between parts that I can't get to mesh continuously, I still have this problem.

Here's what I want: can someone PLEASE recommend a different FEA package that won't give me so much hassle??? I'm frankly sick of COSMOS when it comes to thin shell assemblies.

David

 

[cbrn]

Hi,
unfortunately, meshing is always a hassle. The program may give you more or less control over the mesh so that you can achieve exactly what you need. In my opinion, Ansys is one of the "most powerful" ones (in years of dayly use, only a pair of times it wasn't possible to mesh the model at all, even with full-manual controls, but then we discovered that there were topological problems in the 3D model...) but my opinion is biased, of course. And there is another problem: cost. Ansys is a high-end general-purpose FE system, so the cost is in consequence... But the results are also, fortunately! ;-)

Change-of-system apart, I find a bit strange the tons of troubles you are having: are you sure the 3D geometry is "prepared" consistently for a shell mesh? And remember that shell nodes have 6 DOFs, not only 3, so also rotations may be a problem!

Regards

 

[jmagri]

I have the same problem with CosmosWorks in shells. The problem is that the "Active" side of the shell needs to be connected to active side of the shells of its adjoining parts. This can lead to some Mobius strip types of arrangements. What I have found that helps is to create the shell by surfaces including the sides of the part. This leads to some compromise in the model but will allow the thing to mesh and have the parts stay connected. You can determine this by comparing the mesh on the active or "Orange" side of the model to see that the elements are on different parts have common node points where the edges of the shells meet.

 

[cbrn]

Hi,
yes, this is called "normals matching" and it helps in applying the boundary conditions in the right (i.e. expected) direction(s). Possibly, if you don't check the normals for a set of selected surfaces where pressures or other directional "things" are applied, you'll end up with an ill-posed problem, which will never converge (or, worse, will converge to a solution which is NOT correct !!!).

Regards