FEA of welds, shell vs solid

[kribri]

I have a question regarding the use of shell elements vs solid elements in the FEA of welds.

Some say that the shell elements don't give as good results for the peak stress as the solid elements. But shell give okay results for the nominal stress.

Any comments/experience on this?

K

 

[corus]

Would knowing the peak stress at the weld toe be of any value? Nominal stresses away from the weld, or mean stresses within the weld, are the values used most often.

corus

 

[TGS4]

To add to what corus is saying, can you actually characterize the EXACT geometry of the weld profile (particularly if it is left in the as=welded condition)? How would you deal with undercuts, or lack-of-penetration? How about fillet welds with undefined back-sides?

What is your purpose for wanting to know the stresses at a weld? Fatigue?

 

[kribri]

Thanks for the quick feedback.

I see that I was a bit quick when posting this, sorry for the incon.

I'm not interested in the peak stress in the actual weld, but peak and nominal stresses in the surrounding structure in combination with studying mean stress in the weld.

The question is, based on the above, can this be done by using shell elements? Is this a normal procedure? Any links to FE procedures?


K

 

[ishvaaag]

I think that your question is useful more than anything to those attempting to model in FEM welded unions. In such context one would thing that the more complex the program (allowing for more flexibility in the model conditions) should be better, and so one would tend to discard shell modeling against 3D.

For most others, sticking to union design by more common procedures would seem the way to go.

Then, use of shell, plate elements can have some use to see the stresses in them, say, in a bridge situation you have some other loads applied and try to see if the presence of the weldement would look to give some zone too bad a behaviour. Even for that 3D should do better, anyway, but insight you would have.

Also, FEM results, like opinions of structural engineers, there are of various kinds. I can assert for sure, for I have seen it with my eyes, that the results of plate elements in RISA 3D and one plate in Visual Nastran 4D give different enough results to make some deflection permissible or not, or vary the proposed reinforcement. I am also reasonably sure that both can be used to both intents for I have practiced it and none has come to my door to commplain, but results, are different, and must be, besause the FEM formulations are different. That the difference should be smaller or one should be corrected more than the other to approach some theoretical solution, this is for others to say.