RODs in NASTRAN

[rudiklk]

Hi community,
i have some kind of trouble using a ROD in NASTRAN.
My problem is a completed analysis where i thought a fatal error has to occur.
Its a bench-scale model and is as follows:
A square plate of 6x10 plate elements with isotropic material constrained with SPC DOF 1-6 at one end. The center node of an RBE3 is 10mm above the plate, the guiding nodes attached to the plate. A ROD Element is attached to the center node a tensile force acts in direction of the ROD. The torsional constant J in PROD is 0.
Since the Rod should only be able to carry loads along its axis, the RBE3 center node should act as a ball joint and I expected Rigid Body Modes R1, R2, R3 of the ROD.
I experience this in other solvers.

The center node of the RBE3 is DOF 1-6, but theres a solution even with just DOF 1-3!

Is there any param adding a stiffness to the ROD like K6ROT to plate elements?

Bytheway AUTOSPC is NO.

Can anyone give me some hint?
Thanks in advance
rudi

 

[fkmeyers]

I built your model in NEiNastran and can verify the behavior with AUTOSPC set to YES. We have a demo copy of MSC.Nastran and that also gives the same behavior. It makes sense because AUTOSPC constrains the free end of the rod except in the axial direction. That would prevent the mechanism at that node. With AUTOSPC set to NO in both NEi and MSC Nastran I get a fatal error. What I did notice is with NEiNastran and the iterative solver (PCGLSS) with AUTOSPC set to NO it does not give a fatal error but does give infinite displacements and a bad epsilon. I would consider that as unusable as a fatal. So I do see the mechanism with AUTOSPC set to NO and I see nothing strange happening here.

 

[GregLocock]

Does having a zero stiffness in a linear model sound like a good idea?



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[rudiklk]

Is this of any help to the problem described?

No, zero stiffness is not the problem, with J!=0 Nastran still delivers a result.

I simplified the modell again, no more RBE3, just four plate elements and the rod.

Why does it solve?

PARAM,AUTOSPC,NO
PARAM,MAXRATIO,1.E+8
PARAM,GRDPNT,0
CORD2C 1 0 0. 0. 0. 0. 0. 1.+MSC.VC1
+MSC.VC1 1. 0. 1.
CORD2S 2 0 0. 0. 0. 0. 0. 1.+MSC.VC2
+MSC.VC2 1. 0. 1.
$ MSC.visualNastran for Windows Load Set 1 : NASTRAN 1
FORCE 1 442 0 1. 0. 1000. 0.
$ MSC.visualNastran for Windows Constraint Set 1 : NASTRAN SPC 1
SPC 1 376 123456 0.
SPC 1 378 123456 0.
SPC 1 381 123456 0.
$ MSC.visualNastran for Windows Property 1 : Untitled
PSHELL 1 1 4. 1 1. 1 0.83333 0.+PR 1
+PR 1 -2. 2.
$ MSC.visualNastran for Windows Property 102 : Untitled
PROD 102 1 78.54 10. 0. 0.
$ MSC.visualNastran for Windows Material 1 : Untitled
MAT1 1 210000. 80769.2 0.3 7.85E-9 1.2E-5 0.
GRID 252 0 10. 0. 200. 0
GRID 253 0 0. 0. 200. 0
GRID 254 0 -10. 0. 200. 0
GRID 376 0 -10. -20. 200. 0
GRID 378 0-1.52E-5 -20. 200. 0
GRID 380 0 -10. -10. 200. 0
GRID 381 0 10. -20. 200. 0
GRID 382 0-1.52E-5 -10. 200. 0
GRID 385 0 10. -10. 200. 0
GRID 442 0 0. 30. 200. 0
CQUAD4 61 1 253 382 385 252
CQUAD4 63 1 254 380 382 253
CQUAD4 66 1 382 378 381 385
CQUAD4 71 1 380 376 378 382
CROD 1338 102 253 442

 

[rudiklk]

@fkmeyers

That thing keeps me busy all day!

I downloaded the 300 nodes Demo on NEiNastran and did it there.
What is the iterative solver (PCGLSS)?
I exported the analysis model the way I know as linear static and I experience the same results you decribed,
with AUTOSPC,NO the ROD-node has infinite displacement in all DOFs but the one the load is in.
With AUTOSPC,YES the displacements are 0 and 0.15 in the load direction.
All other results are exactly identical (SPC Forces, Stresses, ...)
And theres no fatal error in the OUT file.

So somehow the NASTRANs can solve the problem even with the redundant DOFs of the ROD. Other solvers like PERMAS stop with fatals.

I found stuff like that in other bench-scale models before, I am just interested in the reasons behind this behaviour.

 

[fkmeyers]

This actually has more to do with AUTOSPC than anything else in my opinion. Other codes may not have such a feature. Now if MSC is solving this with AUTOSPC=NO then you have got me. The only reason why NEiNastran does run with the PCGLSS solver and AUTOSPC=NO is it does not do a full decomposition so the mechanism detection is way less sensitive.

 

[johnhors]

rudiklk - What is the iterative solver (PCGLSS)?

How about "Preconditioned Conjugate Gradient Linear Sparse Solver" ? (just a guess!)