Shell FEA Averaging/Smoothing Across Part Boundaries

[CKerr]

For a shell model (ASME Sect. 8 Div. 2 analysis) of welded structures, when is it okay to average across part boundaries and when should you look at the parts separately? I realize solid sub models should be used to examine the welded areas because of discontinuities, but I was taught that it is okay to average across welded parts because the material is the same and these are full-penetration welds. But I remember reading somewhere that averaging should never be done between welded parts. Thanks

 

[Mustaine3]

My general opinion:
When you see a significant difference between averaged and non-averaged results, then it means that your mesh is not fine enough that area.

 

[FEA way]

You should always check the unaveraged results as part of standard postprocessing verification, not only in case of weldments. Some FEA programs (like Abaqus) even allow you to display discontinuities (differences in values between adjacent elements) in place of standard output variables.