Fine meshes....

[NickE]

I have a small (punny) difficulty, most of our parts are produced (stamped) from coil steel btw 0.003" to 0.040". When meshing in order to solve (the mesh) it is my understanding that the mesh size must be smaller than the smallest dimension of the part. (IE: so that all nodes lie in the body of the part) When I try to do a simple FEA on a part thats 0.012" (w/ mesh size set to 0.012&quot thick it seems that the mesher locks and quits responding. My video card doesnt re-draw well right now so I cant be sure the system is locking, (winXP, P4, 3Ghz, 1G ram) since I cant see the buttons when the CPU cant devote time to graphics processing.

Is it really gonna take >1hr to mesh my very thin parts? I did a 0.040" part the other day and that wasnt too bad even with some sections controlled down to 0.010" mesh. Is there a workaround for very thin stuff?



Nick
I love materials science!

 

[ShaggyPE]

Are you using CosmosXpress or CosmosWorks?
I have only limited background in FEA but I do know that CosmosWorks will allow you to use "shell elements". Shell elements are specifically for thin walled parts. Your results will be more accurate too.

 

[NickE]

CosmosWorks....

Ahhh. I wondered about that, Am I right thinking that shell meshes are created by just meshing in 2-d the surfaces then connecting the surface nodes to create 3-d polygons?

nick

 

[Arlin]

Actually, shell mesh elements are only 2D polygons (no 3D polygons what-so-ever). They are simply a flat polygon (usually a quadrilateral shape) with some sort of thickness property. Of couse, even though the elements are 2D, they can be applied to thin-walled 3D models. A shell mesh is essentially a surface mesh usually placed at the midsuface of the thickness. SWX even has tools to create midsurfaces with wich to mesh.

Think of it this way: a soccer ball looks like a thin-walled sphere meshed with hexogonal elements. Or, to mesh a length of L-shaped angle iron, only 2 rectangular elements would be necessary.

Shell elements are MUCH more efficient and practical when analyzing a thin walled part.

 

[TheTick]

I'm not so sure that rule (mesh size < thickness) applies across the board. Typically, that applies to full 3D meshes, where you want at least 3 distinct layers of elements through a thickness. Not necessarily so with shell elements.

You need to run multiple analyses on progressively finer meshes and check for convergence. When you have successive analyses with nearly identical results, you have found a good mesh size for that case.

Also, try local mesh refinement in areas where the stress change gradient is most extreme (stress concentration areas).

All this machinery making modern music can still be open-hearted.