Does FEM analysis correctly model buckling? 1

[JAE]

For finite element models, say RISA or some other program, where you have a structure under various loadings, does the PDelta analysis performed by these programs properly estimate buckling?

We've been having this disussion in our office and one view is that the PDelta certainly estimates the second order effects, but that this is not the same as Euler buckling.

The other view is that Euler buckling is simply a derivation of second order effects that uses an assumed out-of-plane initial distortion to get second order effects started. So with a finite element model with forces in two or three global directions you will, by nature, have the second order effects started and the buckling load will be at least approximated.

What do you think?

 

[Ron]

ishvaaag...you are correct that there are programs that handle such; however, they are programs that can handle dynamic displacement in a variable stress continuum. Programs such as NASTRAN and similar can do this; however, they are not typical "stick frame" analysis programs and are difficult to work with in building structures (though with patience and a lot of pre-calculation of properties, it can be done).

 

[JAE]

This thread - my original question - has nothing to do with codes..just fyi. I'm not interested in the codes for this question.

FEA analysis does correctly derive the stress levels in odd shapes and structures (assuming small enough elements).

If you take a curved shape - a really significantly curved shape: think half cylinder - and add axial load, the finite element analysis will correctly derive PDelta effects as long as it is in the elastic range.

The question is - does buckling always involve inelastic conditions?

 

[ishvaaag]

Sure, Ron. And for those that maybe someday may want to replicate my model of the last post I forgot to quote that selfweight gets therein factored by 1.35, whereas the quoted additional load gets just 1 as a factor. Irrelevant anyway to the issue.

 

[IDS]

The question is - does buckling always involve inelastic conditions?

No, Euler buckling is based on elastic behaviour. On the other hand in real structures buckling behaviour is often initiated earlier than would be indicated by Euler buckling theory because of inelastic behaviour. This is one area where a correctly set up FEA will often give a more accurate and more conservative result than application of simplified theories in a hand calculation.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[BAretired]

JAE,

If that is the question, the answer is no. Buckling does not always involve inelastic conditions. Long columns buckle elastically. Intermediate columns buckle inelastically. Short columns don't buckle.

The buckling load of a compression member may be governed by flexural buckling, torsional buckling or torsional-flexural buckling.

I thought that the question in this thread was...does RISA or some other program properly estimate buckling? Possibly some do, but apparently RISA does not, at least not at the present time.

BA

 

[JAE]

RISA stays within the elastic condition of the material so my question was related to RISA - but based on the query as to whether inelastic conditions must occur for buckling. BA - your long-intermediate-short brings back some memory of that - I guess I could very well do some research and re-study the Euler approach.

The secondary question, with a finite element model such as the one I posted above, would be - is there any way to tell if you are in a long-intermediate-short condition with an odd, irregular plate layout.

 

[GregLocock]

FEA programs such as LS DYNA are entirely capable of modelling post buckling and post plasticity dynamic behaviour, otherwise they'd be useless for vehicle crash analysis. Typically they use a 1-20mm element size.

I'm a bit staggered to read some of the above comments. It's a big world out there, not everybody gets to design to code.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[SAIL3]

Instead of calling it pl buckling, lets look at it as the effective width of the pl for vert compressive stress.
Take one of the lower panels which looks like 5'wide by,say, 5' high.
A rough guess of the effective width would be, say, 40t to each
corner. Assume 10ga...40x.135=5.4". So one is left with an effecive angle in the corners of 5.4x5.4".
What happened to the rest of the 60" wide pl....did it buckle?
Not in the sense of a col ...it just has no load carrying ability as far as compression is concerned..not a problem as long as the load can get to the corners.
Take shear loads...h/t=60/.135=444...is shear buckling a possibility?..not a problem if one can develop tension field action.
Does one look at the sides as a deep girder, etc.
These are all questions I would ask as I believe that the FEA program is not set up to handle because there are just too many variables for each different case and that only engineering judgement can address.

 

[ishvaaag]

GregLocock, there's sure a lot for all to learn. I would think one shouldn't expect -realistically if not ideally- the same degree of specialzed expertise of one structural aerospace or automotive engineer than of an structural engineer or an architect. I have been always impressed by the extent of mathematics that were retained by those dealing with material strength and elasticity in the early XX century era, and missed too much not to be able to match their stature.

As you say design by the code may become difficult to attain. But real structural designers are demanded quick solutions to structures with not too much time to provide an answer, with many other perentory tasks to do concurrently and still, as usual, finding some difficulty in being economically functional. In short, real engineers, as in the vapor era, make some mistakes and some boilers explode.

It is pertinent to signal that no building is being thrown the dedication and expertise nor in hours nor in demanded technical stature of the intervening professionals than something as the automotive or aerospace or weapons industries is having; nor the price of the items are comparable.

So I instead of becoming baffled am reassured that to get to any level of quality you either throw the required means or you get not, I see logic in it and not bafflement. You get what you are paying for.

 

[dik]

I don't use FEM for general design, but for determining forces and moments; I then use spreadsheets for the actual design check.

There is one exception and that's for curved archrib structures. These are laminated wood, 3 pin arch structures and I have found a close correlation with analysis and design. By making minor changes in the depth of the ribs, you can review the condition as it starts to 'buckle' or have the design moment diagram change 'wildly'.

Canadian codes require these be checked for unbalanced loadings and the effect is quite dramatic. I've used it to show some clients why a change in span or geometry has a large increase in the cross section.

Dik

 

[BAretired]

Greg,

It's a big world out there, not everybody gets to design to code.

Structural engineers are required by law to design to code although some might prefer to do otherwise.

BA

 

[JoshPlumSE]

Some comments about P-Delta analysis:

For frame members, you would normally only consider the lateral joint translation and axial force in the member as a contributor to the P-Delta effect. It is not difficult to modify the local element stiffness matix of the member or introduce secondary shear forces which induce an effect equivalent to the P-Delta moment.

Now, look at a plate element. You've got a two-dimensional element subject to a multi-dimensional stress state. The destabilizing effect of this system is not nearly as clear. That's why most programs do not consider P-Delta effects for plates.

The programs that do include P-Delat for plates may be using an inexact approximation of the effect (this is what RISA does for the wall panel elements). In which case, I don't know that you can expect a P-Delta analysis of a plate element to truly capture the elastic buckling effect in the way that it woudl do so for a frame model..... In my opinion, you'd probably have to do an Eigenvalue buckling analysis to have more confidence that the buckling effects of your plate element model were adequately captured.

 

[TERIO]

Euler buckling is an eigenvalue problem which can certainly be solved by FEM. Some notes on stability analysis in FEM are provided in part IV here:

http://www.colorado.edu/engineering/CAS/courses.d/NFEM.d/Home.html

 

[JAE]

JoshPlum - I just did a quick FEA wall in RISA 3d and voila!!! there is no second order effects with the shell elements.

After reading the help on this I'm a bit surprised as my own PDelta knowledge from past FEA studies - both two node and multi-node elements - the PDelta analysis was performed on the deflections of the joints - and not as RISA does, by adding shears to two-node elements only. I'm not sure I like that (despite liking a lot about RISA).

 

[ishvaaag]

JAE that reminds me the B-Louis some something that "so much touted, never seen".

Running 1 with P-Delta as it is 175 hypothesis in RISA 3D for a Portal Frame Building 25x50 m long takes about 15 mins in my now becoming midrange PC. Now imagine the same for iterative reconstruction of the matrix stiffness in each of the hypotheses and cycles ... no wonder the sellers of structural analysis shy from such approach.

 

[JoshPlumSE]

JAE -

Use the Wall panel element in RISA rather than pure plate elements. Then you should get your P-Delta effects. Though I should point out that it is a P-Big delta effect.

P-Little Delta effects for walls generally have to deal with a code based iterative procedure that considers the cracking of the wall (see masonry and concrete codes). In RISA, this is taken care of the wall force and code check calculations rather than being considered directly in the analysis.