FEM vs. Analytical Methods or Interpretation of FEM Results 2

[mihumihu]

Static analysis using analytical methods is often based on rather crude models of reality - beam theory, net section stresses, no or limited interaction of elements, ....
Still, for ductile metals these methods work well, are widely accepted and give a quite distinct result: MoS > 0 is good, MoS < 0 is bad.

Now, if I run a linear elastic FEM analysis much more details - notch stresses, interaction etc. - are covered and show up in nice red colour on the result plots, and suddenly everybody starts to worry.

How do you interpret FEM plots with peak stresses at stress risers for static analysis? Based on a hard stress limit? Do you say good or bad by engineering judgement? Do you run a elastic-plastic analysis and check for plastic strain limits? Do you use references (which?) or rules of thumb?

Thank you
Michael

 

[IRstuff]

There is a complete forum for FEA questions: forum727

You should probably RF this and repost there.

TTFN

FAQ731-376

 

[SWComposites]

1) junk the FEM and use the classical methods that have worked well for years. FEM is just another (maybe less crude) approximation to reality that generates lots of output for which it is difficult to determine what to do with, and pretty pictures for management.

2) correlate the FEM to the classical methods and test data. Analyze the test articles with the same FEA modelling approach that you plan to use for your actual structure, compare the stress/strain state in the FEM at the failure load to the test results, and determine an appropriate failure criteria, post-processing methods, etc so the FEM analysis matches the test results.

SW

 

[rb1957]

"How do you interpret FEM plots with peak stresses at stress risers for static analysis?"

examine them very carefully ... they are probably very localised, not thru thickness. one "aid for presentation" is to cut-off the fringe plot at yield (so that everything above yield shows red, but at least you don't see the relatively meaningless 500 ksi stress).

another thing to note is that is has been happening all along and if your bosses are only realising this now, 'cause they can see it !, get other bosses !

 

[YoungTurk]

A strong second for SWComposites comments.

Regarding your estimate of classical methods giving "quite distinct results" - keep in mind that those MS results are reached after many many assumptions and approximations. You still need to exercise the same judgement when performing FE analysis.

 

[GregLocock]

One approach has been to reanalyse a known successful part using a new methodology and then set the performance of that part as the new pass criteria. We call that a bogey test. Each new release of software is actually checked on bogey parts to make sure that that is still a valid assumption, which is why we are so many releases behind the current one (typically two years).

Cheers

Greg Locock

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

 

[mihumihu]

Thank you very much for all the replies.

@IRstuff: I chose this forum because I am not interested in the FEM side of things, but in the interpretation of results in an aerospace structures context.

Correlate to tests / reanalyse known parts:
I agree this ist the best way to go, but for modifcations / repairs it may not be feasible due to time and budget restraints.

Use classical methods:
This would probably be the easiest way to go, but still I think FEM has its merits in understanding how the loaded structure really behaves

Cut off at yield and examine carefully:
That´s what I usually do. I accept small high stress areas as long as I can use engineering judgement to explain it is just a notch effect that does not affect a large part of the crossection. Sometimes I read forces on structural parts from FEM and use them for classical analysis. But I wonder if there is a better, more repeatable way to get to a good or bad conclusion.

...has been happening all along and if your bosses are only realising this now, 'cause they can see it !, get other bosses !:
After some laughing I realized the truth in this statement. But in the end it´s not the boss that I am worried about, but about convincing myself, and the certification authority in the end.

To provide some background information: I have a substantiation report from an OEM done the classical way (beam theory, original a/c structure takes all a/c loads, additionally installed structure takes only loads from new installation, no interaction, stiff structure) that shows a certain installation is good. In the FEM model I get high stresses due to the interaction and deformation of original and new installed structure. The high stress area is not just local at a fastener hole, but covers the crossection of a stringer. Still I think (but can´t prove it) that the stringer would not fail but just slightly bend and redistribute the load.

Thank you
Michael

 

[SWComposites]

"The high stress area is not just local at a fastener hole" - how are you modelling the fastener hole? why are you modelling it? Stresses at fastener holes are almost always incorrect values.

"How do you interpret FEM plots with peak stresses at stress risers for static analysis? " - there is no one simple answer, except "it depends". On the particular structure, material, stress location, modelling idealization, etc, etc. Hence the comments in the above posts.

Which way is the stringer being bent? axially or laterally? If it is being bent laterally, then that may be a bad thing for overall panel stability and strength.

It sounds like you should be using the FEM to determine internal load distributions, then should be using those loads with classical type analyses to calculate margins.

SW