I have to check req von misses stresses and total strains of welded model.
B.Conditions are static force on applied in 4 holes 400kg, gravity, 3d fixed constrains on down edge tube. Sheet plates thcks are from 5 to 10 mm.
Mesh should be according to best practice i.e min 3 elements per thck to recive reliable results.
I would like to see any model (movie would be the best) how to prepare 3d welded model with welds, B.C's and proper mesh.
Thank you very much. My model is attached.
Solved! Go to Solution.
Yes, some files are missing, but we can work from simulation geometry. In this case I suggest using shell elements instead of volume elements because these parts have thin walls. I've replaced weld beams by rigid connections to representing relations between parts.
Here is a video about steps:
Thank you Sir for this complex answer! One question beacuse Im not sure now.
You have used shell element. Is it proper to use shell elements for plates with 10-12 mm thickness? If yes, please why more deeply.
In the past, I have used shell elem's only for shell plates with thickness less than 2 mm rather than solid elements for thicker plates...
Best regards, CB
Shell elements should be used if the geometry is "thin-walled". In other words, if the overall dimensions are very large in relation to the thickness of of the geometry. For example, you might have a 10 m x 10 m hopper with 20 mm plate walls. If you tried to mesh that with 3D elements the model would be too big to solve because you would need at least 3 - 5 layers of elements through the thickness to get any meaningful answers. So instead, use shell elements. It does not matter about the thickness.
Thank you Sir, more clear now. I think such model wont be so big since 20 milion D.O.Freedom model is not such a large model rather than 10 years ago, right? Anyway you are correct.
Is shell element in SE still valid for larger deflections?
Thanks for any detailed answer!
Twenty million degrees of freedom? If there are 5 degrees of freedom per node then that is something like 4 million nodes. I don't know about you but a model of that size would kill the budget on most projects around here especially if multiple runs / iterations / non-linear / contacts are needed. That's assuming one had enough processors / RAM on hand. If it was me I would seriously use shell elements. Four million nodes could turn into a few hundred thousand or less.
Just a suggestion, but this is the kind of conversation that might get more attention over on the FEMAP Forum. You might want to click on that link and give it a try.