Nastran sol 106 & BUCKLING

[Trajano]

Hello

I have a FE model with non linear material properties.
If I run a SOL 106, will I be able to see the buckling modes?

Tips:

1. There is also a method (that implies the use of PARAM,BUCKLING among other things) to calculate the buckling modes. But I´m not talking about that, just about a simple SOL 106.

2. I´ve seen some models where they put certain nodes out of its plane to induce the buckling.

And finally. In case that I can see the buckling with the SOL 106, why I cannot see it with the linear SOL 101?

 

[kellnerp]

You will just see the first mode since it is not an eigenvalue solution. You can also use a small out of plane force to excite buckling behavior. You might try different patterns of small forces to excite different "modes". Buckling can be rate dependent too. I am not sure that SOL 106 will pick this up.

TOP
CSWP
BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."

 

[drk186]

I can't remember what SOL 106 is (non-linear buckling, right?)

If that's the case, you should be able to get your mode shapes from it.

Moving nodes certainly can introduce a moment into your model and induce buckling.

SOL101 is linear static solution, and does not have any eigenvalue extraction parameters, which you need for a buckling analysis (linear or nonlinear).

 

[Trajano]

Thank you both for the replies.

drk186, sol 106 is the nonlinear or linear static. And, please correct me if I´m worng, but I think that if I don´t use the PARAM,BUCKLING, the SOL 106 doesn´t calculate any eigenvalue. Hence, my problem is how can I tell if certain part of the structure has buckled or it is just bending. Of course, the first evident answer is to run an buckling analysis (waht I already did in my FEM), but I´m just trying to understand the way that SOL 106 shows (or doesn´t show) the buckling.

And yes, kellnerp, I guess that if I can see a bucking mode it will be the first mode. I think I like better the idea of using out of plane forces than moving the nodes.

Cheers

 

[kellnerp]

If you do a linear buckling solution on the model, how close together are the solutions?

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."

 

[Trajano]

kellnerp, the shape of the buckled structure resulting from the linear buckling eigenvalue solution is the same as the one from the nonlinear static solution (of course in the nonlinear one appart from the bucking there are other deformations in other zones that have nothing to do with he buckling). And it happens around the same level of load that the linear solution says (I say "around" because it is difficult say exactly when happened, since the rest of the structure takes the post-buckling loads)

 

[kellnerp]

OK, so now it gets interesting. Obviously the linear solution will tend to give a higher buckling load than real life. So you have to apply some kind of knock down factor. Is this a shell model? If so you might be able to introduce some randomness to the structure to simulate "as built". This will lower the eigenvalues and the non-linear solution as well.

Are you trying to get the post buckling behavior?

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."

 

[ThomasH]

Hi

I think this is the way it works:

SOL 101 is linear buckling. Basically it computes eigenvalues for the undeformed structure.

SOL 106 is a nonlinear solution with or without buckling. If you include deformations in the solution (LGDISP) the solver will for a large load have trouble with the convergence. That I think will be reported as structural failure but doesn't neccesarily mean buckling.
Now, if you instead use PARAM,BUCKLING for the nonlinear solution you will get eigenvalues for the deformed structure. And if you exclude the deformations from the nonlinear solution (LGDISP=0) you'll get the linear solution. Provided there are no other nonlinear effects included.

As for real life, it's nonlinear but also nonperfect. A straight bar isn't absolutely straight and probably also has some initial stresses. These initial conditions can to some extent be included in the nonlinear solution. But that depends on what they are and also if the imperfections are known or if you have to make assumptions.

Maybe this helped to some extent.

Good luck

Thomas

 

[kellnerp]

Trajano,

When I said, "See how close the solutions are." I meant the eigenvalue results of the SOL101. If the eigenvalues are far apart set your forces or node displacements at the displacement peak locations of the corresponding mode shapes of the linear run to excite different modes in your non-linear run.



TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

"Node news is good news."

 

[Trajano]

Thomas, your explanation is very clear and useful, thank you.

Kellnerp, thanks again for your posts. I misundestood your question about "how close the solutions are". In fact the first eigenvalues are quite close (0.987; 1.033; 1.042; 1.083).

My model is a section of a Aluminium longeron. It is basically a web reinforced with horizontal and vertical stiffeners. For the studied load case, the instability is in fact the local crippling of one of the stiffeners, but the surrounding structure is able to carry the extra load and in fact the buckled stiffener doesn´t even yield at limit load. I have no problems at all with that crippling (not even convergence problems), but it made me wonder a general question about the capacity of the SOL 106 to show me the bucklings (or mine to tell if what I see means that the structure is buckling). And
if it has that capacity, why do we need to calculate the eigenvalues with PARAM,BUCKLING?.

But after thanks to this thread, I´ve come to the conclusion that for a complex structure you will never be sure that the SOL 106 (without PARAM,BUCKLING) is showing you the buckling and therefore an eigenvalue analysis has to be performed.