Unexpected phenomena at end of beams

cashan

He tries to carry out a simple analysis of the crane foundation. Everything comes out rationally, except nodes, where the joists are connected to the shell (I use high stiffness beams to apply a moment at the top of the epedestal). It looks like the nodes at the ends of the beam have not rotated with the beam. Can anyone explain this phenomenon or say how else to add the moment at the top of the column?

Victor

I don't know why it seems to be bending the opposite way there.

Another way is to apply the moment load to a ring of faces around the top and it'll be distributed as equivalent forces. It doesn't work reliably with shell elements and CCX but you can add a layer of dummy solid elements to carry the load on their faces that are coincident with the shells.

There's a lot of extra deformation of the pipe without the stiffening of the beam elements but you could increase the E of the solid ring to regain that if you want it.

cwharpe

What also works to retain the top-edge shape is to increase the stiffness of your spider beams, whether by bumping the spider's Young's Modulus up by, say, x1000, or by increasing the spider cross-sectional inertia by a similar magnitude. I seem to recall one author preferred to "fill-in the area" the spiders spanned.

I've performed segmented beam comparison tests with Hollow Structural Sections (eg. HSS 4x4x1/4"), formed of shell elements, and replaced the middle-lengths with beam elements using just the same spider-beam diaphram interface as you show. The deflection & moment profiles were within acceptable limits comparing between theory, continuous & segmented cross-sections for varying end-conditions. Conclusion: Values near the spider boundary/load interfaces might be suspicious or even look odd, but downstream results away from the affected zone should be good.

@Victor: It is suspicious to note what happens at the spider beam-ends. That rotation and expansion of the end-faces resemble the hourglassing behavior seen elsewhere. The phenomenon seems to attenuate when the stiffness is adjusted higher, but maybe only relatively so.

JohnM

Try node-surface coupling

Victor

@cwharpe I see your suggestion of increasing the beam stiffness addresses the underlying problem - it was already qualitatively correct but for very flexible beams! By refining the beams, you can see why the lip bends backwards and why the beams rotate (and expand due to exaggerated display scaled factor). The rotation isn't consistent with the shell rotation but I think that's because the shells' lack of a drilling DOF so the beams have freedom to rotate normal to the surface.



The only unexplained effect I see is that alternate beams in the unrefined model rotate much more than others. Perhaps that's due to hourglassing??

Moral of the story - mesh refinement solves surprisingly many problems!

cashan

Hi,

Thank you for all respond!
"node-surface coupling" looks good. However, I have a problem with using it. Can you upload a ".lim" model so I can see how it is added? Or maybe there is a tutorial on the web?

JohnM

Here you go. As Victor suggested, I added a small ring of shells just above the pipe and made them my surface. Select that surface, switch to "nodes connected to that surface" and use the "insert node between" to get a center node. Name both the faces and master node, then open up the node-surface-coupling option in loads and constraints.

cashan

Thanks! It's easier than I thought! Works great.