Local buckling in bending and Inventor Nastran 3

[TimSchrader2]

Hello

Anyone know how accurate the Inventor Nastran program is in predicting local buckling. The AISC (and other)guidlines on b/t ratios and reduced allowable stressses would be the first trial. But as the below thread notes local buckling is almost always underestimated by a large margin (per Timoshenko). So I will use the above program to further define these limits.


Designing a tri telescopic boom with welded Box beams, but with a slot down the center of the bottom plate. So really a open beam. With some HSS, which could make the problem even more critical. But the HSS MFG's (SSB) says they do to what they call "stress redistribution", so the HSS steels handle local buckling better then regular steels with the same E.

This could be a big $ question.

https://www.eng-tips.com/viewthread.cfm?qid=268445

 

[BAretired]

But the HSS MFG's (SSB) says they do to what they call "stress redistribution", so the HSS steels handle local buckling better then regular steels with the same E.

Do you believe them?

BA

 

[TimSchrader2]

Hello BA. Not fully, but not sure how much. So I posted this

 

[BAretired]

Hello Tim. I interpret their comment to mean they remove some of the residual stress which occurred during the cooling of the shape, sort of like annealing. I'm not sure how effective that would be and would like to see tests demonstrating an improvement on local buckling but I am no expert on that subject.

BA

 

[human909]

Not directly answer the question but following on the cantilever buckling discussion topic had a few weeks ago I've been playing around with NASTRAN a fair bit recently and trying to get a handle is ability to model and predict buckling. I've been happy with its predictive results but none of that is local buckling or compared against actual results.

One way to approach your question is to test or get data from somebody else's tests (eg academic papers) and compare it against NASTRAN results.

 

[ThomasH]

TimSchrader2
Are you talking about the linear buckling in Inventor Nastran compared to a local buckling in a HSS section?

When you say Inventor Nastran I assume that you mean the software that was known as Autodesk Nastran and before that NEiNastran? If it is then the analysis result is comparable to Euler buckling for columns. That means that neither the material strength nor any imperfections or residual stresses are included. It is an "ideal" system.

Provided that you have a good mesh I think Nastran will give you a reliable result for the linear buckling. But you can't use the result for design.

Thomas

 

[FEA way]

Inventor Nastran’s solver is very good. But don’t trust the linear buckling analysis in actual design, as ThomasH said. It may underestimate critical load on dangerous level. This is an issue with all linear buckling analyses, not only in this program. I’ve seen some examples where results from such simulations were pretty much useless wen compared with correct ones. However you can try to perform much more realistic nonlinear buckling analysis. To do this you will have to go for nonlinear structural analysis with imperfections imported from previous linear buckling analysis. It won’t be easy to set up in this program but should be possible.

 

[TimSchrader2]

Hello
Inventor Nastran does also have the non-linear analysis, which I have started to work with. Its a bit trickyier then Fusion 360.

Yes, I am specifically talking about Local buckling of a HSS section as predicted by Inventor Nastran.

I will start the design with the d/t ratios from the AISC code, then run the Buckling programs. But I would not go thinner then the code suggests for d/t ratios. So running the programs is more of a check.

Thanks for the feedback

 

[WARose]

I'm not sure I trust any program when it comes to buckling (for something that isn't a rolled shape; and even then it depends on the shape). There is just too much that can get screwed up (especially when we start talking about slender elements).

 

[BAretired]

I agree with WARose. Practical results for the buckling of a thin walled cylinder don't agree with theory (by a large factor) so what would the program be based on?

BA

 

[human909]

I agree with WARose. Practical results for the buckling of a thin walled cylinder don't agree with theory (by a large factor) so what would the program be based on?

You got me curious so I stared to do some digging... From my superficial digging your assertion about 'thin walled cylinders" doesn't seem to apply to typical HSS which seems is a long way from being 'thin walled'. And yes there are models that closely represent thin walled buckling:

One example, (I'm sure there are better papers):

https://pdfs.semanticscholar.org/803b/233d056c94e06e2305916f9909375cec41e3.pdf

Though I'd happily be shown further that I'm wrong. Like many computer tools FEA anaylsis can be amazingly powerful but can be misleading and dangerous with the wrong assumptions and interpretation.

 

[canwesteng]

I think slender elements is all most FEA can do reliably in the first place. Anything compact will buckle inelastically and if you're using FEA for this you better have a beefy factor of safety. OP - To my knowledge NASTRAN will not to inelastic buckling, and local buckling has almost always been inelastic in my experience (how often do you see a buckled web pop back?). You will get at least an upper bound on the strength of the member, and that can be valuable, but for design you need a reasonable lower bound (likely the AISC b/t ratios).

 

[FEA way]

Indeed this article does a good job showing the difference between linear buckling prediction with FEA and actual nonlinear buckling response. However it’s important to remember that linear (eigenvalue) buckling analyses are just one (highly approximate) way to solve buckling problems in FEA. Assumptions in this case are: linear elasticity, small changes in geometry (stiff structures) and no imperfection sensitivity. The last one is crucial as many structures (such as the shell from article attached by human909) are very sensitive to imperfections. Try to solve classical benchmark problem called Lee’s frame using different methods (linear/nonlinear, with and without imperfections) and you will see the differences.

It’s also important to remember that many FEA programs can perform nonlinear buckling (also called post-buckling) analyses. But there’s often no separate simulation type for this purpose so it may cause the confusion. Usually such simulations are done using general nonlinear static or dynamic solver with geometric ninlinearity turned on. The only difficulty is that not all FEA programs allow users to import mode shapes from linear buckling to serve as imperfections. And even if they do it’s often vey unintuitive. This is the case with Inventor Nastran too.
There’s also a special method for such problems featured in selected software. It’s called Riks (arc-length) method. You can find it for example in SolidWorks Simulation or Abaqus.

 

[WARose]

However it’s important to remember that linear (eigenvalue) buckling analyses are just one (highly approximate) way to solve buckling problems in FEA. Assumptions in this case are: linear elasticity, small changes in geometry (stiff structures) and no imperfection sensitivity. The last one is crucial as many structures (such as the shell from article attached by human909) are very sensitive to imperfections. Try to solve classical benchmark problem called Lee’s frame using different methods (linear/nonlinear, with and without imperfections) and you will see the differences.

Absolutely right. It's great to have a program that spits a buckling value out at you.....but there is a reason virtually no seasoned engineer will run with that without some sort of code or well established criteria to back it up.

 

[tmschrader]

Hello
FEA way. Yes, That does sound the way Inventor Nastran works with a nonlinear turn on. The “acad expert” says it will do non linear buckling and there are some threads in the knowledge base. We Have a subscription so I can talk to them on the phone when we wish. But Going beyond the code guidelines would put me in no man’s land and it’s not a place I want to go.
The other issue with thin wall or high b/t ratios is you have to decipher the aisc additional appendix rules for reduced allowables based on lower effective widths. Which I am in the process of doing now. These seem to vary from edition to edition. Also have A PDF download from SSB HSS people gave me. Will see what happens on how much of the extra yield strength you can actually use. And then if I can convince the owner to accept those results.

Thanks again

 

[ThomasH]

You can find Riks arc-length method in Nastran. I think Inventor Nastran today is what used to be NEiNastran and I used it a few years ago, it worked fine. But Autodesk may have limited the possibilities to set up a non-linear solution. The comment "Nastran works with nonlinear turn on" makes me suspect that. The setup for a non-linear analysis often has a number of options to use.

Linear buckling, I would compare it to Euler buckling for a member. It actually is Euler buckling but solver numerically. It can be used to calculate the buckling for a member in a structure. But it doesn't include imperfections, the materials yield strength and the structures deformation during the buckling. So you can't use the result directly for design, it overestimated the capacity compared to the real structure.

However, the lowest linear buckling shape is in my experience a good start for a non-linear buckling analysis. Use it as base for initial imperfections, include the materials yield strength and large deformations. Then you can have a useful result.

Now, if Inventor Nastran can solve this today, I don't know. NEiNastran together with Femap could do it, that I do know. The same is also valid for several other FEM-solvers, it's non-linear but I would not call it extrem.

Thomas

 

[FEA way]

Inventor Nastran is former Nastran In-CAD and yes, it looks like there is an option for arc length method in this software. In the Nonlinear Setup one should find an option to turn arc length on. However standard Newton-Raphson solver should handle the problem as well.

 

[tmschrader]

Hello


Note that I checked out the allowables stresses based on the AISC rules and found that not much extra allowable stress is given for higher yields like in HSS or even ultra HSS. When the b/t's are high and the section is now a slender section, with thinner walls. I was wondering if the European codes allow more. Since this question on AISC codes is not directly related to buckling programs, I moved it to the code section forum.

https://www.eng-tips.com/viewthread.cfm?qid=457010&action=updated.

Note the transfer from linear to non linear analysis is more then just one pulldown click. Just to clarify my statement above. I have only used it a few times and I am still practising with the Nastran linear and non linear. The fusion 360 simulations are easier but not as specific.


Thanks again for the feedback.

 

[ThomasH]

FEA way
Yes, and NEiSoftware had a product called Nastran In-Cad, the solver inside was NEiNastran. The name has changed a few times. I worked with NEiNastran for several years and it was a good solver.

I don't know how the solver inside Invertor is documented and unfortunately NEiSoftware's site no longer exists. But on that site there was information how to run different types of analysis's. The focus from NEiSoftware was to provide an excellent solver. Their support was also very good. But, to be fair, I don't use that solver today. I still use Nastran but a different prefix .

Thomas