Nastran/Abaqus shell stiffness discrepancy

[derickdfc]

Good Morning Gentlemen,


Just for information I thought you may find the following numbers
interesting.
The model was a gently curved bonnet and was in
combined torsion and bending. Max displacement is reported. The adhesive

referred to sticks the top and bottom of the bonnet together but they
are also welded all around their edges.

ABAQUS S4/S3 44.9 mm
ABAQUS S4R/S3 46.6
NASTRAN CQUAD4/CTRIA3 66.1
NASTRAN CQUADR/CTRIAR 45.7
NASTRAN CQUAD8 70.8
MASTER SERIES 50.1

No adhesive results

ABAQUS S4/S3 107
NASTRAN CQUAD4/CTRIA3 123
NASTRAN CQUADR /CTRIAR 109
MASTER SERIES 114

So the CQUADR appears to be MSC's attempt to match the HKS element
formulation. The CQUADR adds a fictitous stiffness to dof 6 for in-plane
rotational stiffness. According to MSC it's a better element than the
CQUAD4 for planar structures under in-plane loading, and also when the
element shape deviates from a perfect square.

Has anyone got any comments on the above? I regard the ABAQUS results to be the correct ones, but we have a long history of using NASTRAN. These results cause a major problem in comparing 'legacy' analyses with current ones.

 

[BobakaRacerX]

You have very succinctly brought to light one of NASTRAN's shortcomings. I don't have time for a longer explanation (I'm hungry and I'm ready to go home), but I saw your post and didn't want to leave you hanging. This 6th DOF, namely the rotation about the normal to the Shell surface, and the manner in which it enters the constitutive relation is why all Shell Elements are not interchangeable. You can generally reconcile the difference by adding a Constraint Eq., but this will generally make the solver less efficient.

Bottom line: I haven't used NASTRAN or ABAQUS lately (can't afford them), so my ability to compare them may be outdated, but back when I was using them regularly, it was my opinion that ABAQUS was hands down the better code, for accuracy of solution, anyway. So, since it appears that you have both codes at your disposal (lucky you), I would suggest that you use ABAQUS and save the headache.

Of course, this is coming from someone that had to put up with NASTRAN before they got around to adding the "fixed-up" Shell Element! I hope this was of some help. Good Luck.

Bob

 

[izax1]

Just a hint!

I see that you are using both Quad4 and Tria3 in your study. The TRIA3 element in NASTRAN is unrealisticly stiff.

NASTRAN is also defining the type of 2D element you are simulating by defining properties, as opposed to ABAQUS where you select the element name. So for a comparison, it is important that you are sure that you have defined the correct properties in PSHELL.

Bernt