Surface Polishing Load - Nastran 1

[4loren]

In nastran what is the best way to apply a torque along Z axis over a set of shell elements lying in the X-Y. In a simplified example the torsion test will consist of a hollow box. The top of the box (lying in X-Y plane) is essentially glued to a rotating pad. The bottom of the box is constrained in translation.

I assume the best way to apply the torque over the surface is to use a RBE2 and applying a Z moment at the independent node. Is this correct, or do you have competing suggestions? Thanks in advance.

 

[GregLocock]

That's not physically correct is it?

I'd model the interface to the pad using compliant elements to represent the glue, and then make some disgusting approximations on the far side of the compliant layer. Otherwise your load distribution is going to be completely wrong.

Are you interested in local stresses, or general displacements?



Cheers

Greg Locock

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

 

[dmacx]

An RBE3 works well for distributing loads at a point over a surface. The reference grid point would have DOF 1-6 and the averaging grid points DOF 1-3. The loads are applied at the single referencing grid. Are you using FEMAP to set this up?

 

[GregLocock]

Does a forest of RBE3s truly represent how a torque is transmitted into an area? I know how to make that work, but it is more trouble than it is worth. How much torque does the the RBE3 in the centre of the pad transmit?



Cheers

Greg Locock

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

 

[johnhors]

As Greg says rigid body elements should be avoided. Torque can be applied as surface traction pressures. Rigid body elements have no place in real world models.

 

[dmacx]

An RBE3 in Nastran is an interpolation element, not a rigid element so there is nothing to be "avoided". The statement that "rigid elements" should be avoided is also misleading since there are plenty of applications for rigid elements in FEA. Interpolation elements such an RBE3 are frequently used in Nastran. They do NOT add any stiffness. They are typically used to distribute loads and mass.

The manner in which the forces are distributed is analogous to the classical bolt pattern analysis. Forces and moments at the reference grid are distributed to a set of points or nodes based on geometry. A torque is distributed as forces which are proportional to their distance from the center of rotation.

As for the comment "rigid body elements have no place in real world analysis" I beg to differ. First, we are talking about rigid elements and interpolation elements which are used in the majority of aerospace and automotive models built using Nastran. Most of the major automotive and aerospace manufacturers use Nastran directly or indirectly.

 

[4loren]

First off, thank you to everyone and their speedy replys. I am trying to find local stresses in the many layers parallel and below the top of the box. The actual model is somewhat complex. And lastly, what is FEMAP?

Thank you!

 

[GregLocock]

dmacx - most of the FEA problems I see are due to misuse of RBEs.

Cheers

Greg Locock

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

 

[dmacx]

That is not a valid reason for not using them.

 

[johnhors]

dmacx - Like Greg I see many, many Nastran models which use RBE2's and RBE3's for loading and supporting that are quite simply incorrect and produce false results as a consequence. I've lost count of the number of times I've had to prove my point by removing the RBE's and apply distributed pressure loading instead to produce an overall load balance in the model so that minimal supports can be used. Once this is done the results match photo-elastic and strain gauge data to within a few percent.

In unskilled hands RBE's are very dangerous animals ! and unfortunately 4loren doesn't know that FEMAP is most likely the best known PC based FE pre and post processor in the world and thus probably falls into this category !

 

[GregLocock]

dmacx, OK, I see where you are coming from.

However.

In this case imagine we have a grid of 16 equal spaced nodes that represent a square surface, through which a torque is being applied normal to the surface.

You are proposing that the analyst set up a set of rbe's in order to apply the torque.

Where would you put the rbes?

What torque or force do you apply to each of the rbes?

I'm sorry, I don't think the rbes have simplified anything.



Cheers

Greg Locock

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

 

[dmacx]

The single reference grid point or node for the RBE3 is where the load is applied and that is simply positioned in space where it would be in the actual structure. The averaging grid points are those 16 other points you are mentioning where the load is to be distributed to.

I would like Johnhors to simply tell me what an RBE3 does. I know that many in my field do not know and therefore never use it. It is however a very powerful tool. Do either of you use Nastran and if so how do you get around using rigid elements then?

 

[GregLocock]

I use them, very very carefully. Typically I will isolate them from the structure I am interested in by using a transition from the realistic model to the unrealistic loading condition, either by modelling a lot more of the structure and ignoring high stresses near the rbes (st venants), or by carefully arranging the dofs etc so that the artificial stiffening of the structure is minimised, or by using them and then getting the test guys to weld on a huge steel plate onto the real structure for my correlation runs.

Given that 90% of the models I see have frequencies that are high compared with the real structure, I am sure that other people don't put as much time into this sort of thing as I do.

Your proposed usage of RBEs would not put the force into the model in a representatitve fashion, locally. You have welded a thick steel plate to the structure and applied a torque to that. That may be acceptable. I'd try not to do it.

Cheers

Greg Locock

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

 

[johnhors]

dmacx - Yes I use Nastran (and Abaqus and Lusas) on very complex solids with equally complex loading conditions, but at the same time the loading is also statically determinate. I can thus precisely specify uniform internal pressures within pressure vessels, apply variable lug loading pressures using empirical formulas and other loaded surfaces with pressures and still achieve a near perfect load and moment balance. All loading is made as life like as possible. There isn't a part of the structure that is not of interest so the rule of St. Venant cannot be legitimately invoked and neither can using RBE's.