Shell FE with RBE2: comparison of different designs 1

[vultur77]

Hello everybody.

I have a shell FE model of a structure that it is made up of a truncated cone shape (just for the sake of clarity, imagine an ice-cream cone without the tip) and a bottom cap (which has a different taper angle, as it would belong to a bigger ice-cream cone) and I am investigating two different designs that differs only in the upper interface of the truncated cone shape:
Design 1) has a C-section flange
Design 2) has a half T section (i.e., like the capital Greek gamma letter)

I have performed a stiffness/flexibility analysis of the two designs by simply applying an unit load at the bottom interface node of the cap (the upper interface is fully clamped: 123456) and design 2 is 3% more stiff than design 1 in axial direction (cone axis).

If I replace, for both designs, the lower part of the cone with an RBE2 (keeping the bottom interface node and transferring only the translational dofs) I got the opposite: design 1 is more stiff than design 2 (still about 3 %).

This last results puzzles me because I am expecting to have the same behaviour as that I got without the RBE2 (i.e., design 2 more stiff).

I have checked the FE model for errors but it seems everything is ok.
Is there someone who has faced a similar situation who can give me some hints on how to explain this or, maybe, to build up a simpler model to better understand the behaviour of the structure?

Thank you in advance.

 

[rb1957]

isn't 2-3% just noise ?

 

[SWComposites]

Rigid elements are infinitely stiff and in no way simulate reality. The model with the RBE2 is probably meaningless.

 

[vultur77]

Thank you rb1957 and SWComposites for your replies.
You are both rigth and I also reached your conclusions time ago.
The difference is not so relevant and the model with the RBE2 does not better simulate the reality of the problem.
I was just asking myself why Nastran gives me such results and if there was something I could do to check (additionally) for mistakes in the model because I have to use it for further analyses. If I find an explanation I will keep the post updated.

Ps: maybe it's a trivial question but how can I be sure that the difference of 2-3% is numerical noise?

 

[ESPcomposites]

I don't think rb1957 is stating this is numerical noise, but rather noise as far as an engineers expectation. Engineering accuracy is usually closer to the 10-15% range. Beyond this, variances in load, material properties, geometry, etc. are more significant than the analytical differences.

Brian

http://www.espcomposites.com

 

[vultur77]

Thank you for your insight, Brian.
I see it. I am carring out a linear static analysis, I haven't investigated the vibration behaviour of the structure yet. I will try to make some checks in this direction.

Ps: I have re-run the analysis with another version of the software: the results are the same. I keep on checking.

 

[BlasMolero]

Dear Vultur77,
The RBE2 element provides a very convenient tool for rigidly connecting the same components of several grid points together, but should be used properly. Remember that an RBE2 element is not a finite element, but a set of equations that define a kinematic relationship between different displacements.

It depends the DOF coupled internally in your RBE2 elements, but if you consider the joint is too stiff please use RBE3 elements instead and compare result: I think of RBE2’s as displacement links and RBE3’s as force links.

The RBE3 element in NX NASTRAN is a powerful tool for distributing applied loads and mass in a model. Unlike the RBE2 element, the RBE3 doesn’t add additional stiffness to your structure. Forces and moments applied to reference points are distributed to a set of independent degrees of freedom based on the RBE3 geometry and local weight factors.

Best regards,
Blas.

~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director

IBERISA
48011 BILBAO (SPAIN)
WEB:

http://www.iberisa.com

Blog de FEMAP & NX Nastran:

http://iberisa.wordpress.com/