How can I self-learn finite element analysis in a group that does minimal FEM work? 6

[EdwardNigma]

Hi all. As the subject states, I'm interested in learning FEA at work even if it's for simple applications. I currently work in a large stress engineering group where we check suppliers' FEM analyses for aircraft interiors but it comes in the form of a report. We perform simple hand calculations to validate the FEM analysis. However, there are a few people in my group who do model and analyze things from scratch using Catia and Patran/Nastran. Even for them, it's not a full-time job but they model different systems throughout the year. I don't work directly with them. On a day-to-day basis, I typically perform analysis using Excel templates for new designs, capture suppliers' analysis in a report for release, and also perform stress analysis on MRB issues (repairs mainly through reinforcing doublers).

Recently, I expressed interest to my manager about learning FEM, in particular FEA using Patran/Nastran. I listed it on my development plan and he paired me up with a lead stress engineer working in my group on a different aircraft model that does occasional FEM work. The plan as of now is for him to assign me some projects that were done in the past to get my feet wet in modeling and running analysis. Most are not too complex but it's modeling a system, of mostly composite structures, and applying certain loading conditions. However, although my mentor is helpful, he's there to help when I have questions but doesn't have time to hold my hand during this process.

A little about myself: when I first started my career in the aerospace industry, I started as a stress engineer doing extensive static and dynamic analysis using Patran/Nastran. I left job after 2 years to work for a large OEM but switched over to design engineering. Over time, I got the opportunity to work on stress engineering again but it was limited to hand calcs (static and fatigue analysis) of primary structures.

The question I have is, what would you recommend I do to self-learn FEA using Patran/Nastran? Are there books that you guys recommend? What advice would you give me considering my situation? What steps of progression would you recommend (ie tutorials, replicating things I analyze by hand, helping other groups). Are there certain guidelines to follow when deciding what modeling techniques to use?

I want to tell my mentor that I want to work on x, y, and z projects but I need to do some of that work myself and be proactive. I have access to the MSC Patran/Nastran course notes but can't take a one-week course with them because my manager doesn't think it's necessary due to the current statement of work. I found it easy to find information on modeling isotropic materials but not so much on composite materials. I am aware that it's good to go straight to Nastran input files and edit that on a text editor before looking at the results in Patran. I just haven't done it in a long time and would need to re-learn some of this again (debugging text files).

 

[rb1957]

I get this (about FEMs) and GIGO and people not understanding garbage when they see it ("the 'puter said so").

But I contrast this to my experience with CFD. Admittedly our CFD guy is very experienced and we've never had issues, but his "validation" is "I've been modelling for years, applying the same modelling philosophy on multiple projects, and never had an issue". My sense (not being an expert in the field) is that CFD is just as vulnerable to GIGO as FEA, and most of my FEA these days is very simple stuff (momuments, small pressure vessel penetrations) and I still get this validation question. I've managed this with some "nonsense" that people buy, or occasionally by test (9g fwd is simple enough).

sigh

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[Ng2020]

The FAA "generic issue paper" for
FE validation is oft quoted in Safarian's various presentations. Even EASA have referenced it in proposed CM-S-14

Has anyone here ever seen it? I have looked high and low to no avail...

 

[rb1957]

@Ng ... see Sparweb's post above, I downloaded from there.

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[WKTaylor]

SW... Let's see if I can say this without getting red-flagged...

Early-on, Abbott's monthly newsletters were packed with interesting glimpses at some of his works-in-progress... and made for good/thoughtful reads. I, also, [still] like the Abbott website for the abundance of technical data it contains... which makes it a useful resource.

However, over the last +5-years... especially Covid19 years... I have seen that his mind-set and his newsletters have gone-thru a profound shift from straight-forward technical content, to fairly-far-right-wing 'dug-in on his technical opinions inter-mixed with his political perspectives'. Along with this mind-shift... interestingly/oddly... he apparently MOVED his company lock-stock-and-barrel at least 3-times... from Canada to a Caribbean Island and now Britain... I think.

Soooooo... I have actually been watching the tenor of his narrative with amusement and incredulity... especially him wanting certifying organizations like the FAA, TCCA, JAR, etc... to pull their heads-out of their collective anal passages... and get-out-of the way so that real progress would be possible by the smarter people doing real design/development work... like him.

When I posted the link to Abbott's on-line newsletter [this topic] it was actually a loaded question. My personal counter question to Abbott's question/perspective of 'how much oversight is too much'... it what is/would-be the effect of 'too little oversight'. BUT sending him an email on concerned-perspective, I found-out, usually resulted in a reply of... 'so-what'!?!? You don't have to read my newsletter'!

NOTES1. I also post on homebuiltairplanes.com: too much of what 'I see' is based on pretty sketchy understanding of aircraft structural/mechanical design, aviation/flying, history, etc... make for some pretty hairy stuff. I try to help make sense of it all... but a few of the conversations are sometimes too far-out-of-bounds to counter-reply. Sadly.

My 'mentor' was the designer of my dad's T18 homebuilt airplane... and professional engineer... John Thorp. He framed the profession of engineering... and aviation/flying, too... as a total package... in the perspective of "this is very serious and demanding business... that comes with deep personal rewards and intense satisfaction as a career. It is NOT for those who prefer the amateur perspective." Thank You JT!

Regards, Wil Taylor
o Trust - But Verify!
o For those who believe, no proof is required; for those who cannot believe, no proof is possible. [variation, Stuart Chase]
o Unfortunately, in science what You 'believe' is irrelevant. ["Orion", HBA forum]
o Only fools and charlatans know everything and understand everything." -Anton Chekhov

 

[GregLocock]

The counter to the oversight of FEA by bureaucrats is that automotive FEA is ahead of aerospace FEA in some respects (our load cases often involve hard things hitting hard things very hard) to the extent that crash, including airbags, or square edged potholes, or fatigue life durability on 4" chuckholes, as examples, are pretty much solved, albeit computationally expensive, problems. Yet there is no external oversight. Of course the only regulatory one of those examples is crash, and that is checked by physically crashing a car. So in rb1957's example of a 9g impact on a seat, you'd be setting up a rig test on a sufficiently large section of the floor and ripping the seat out. Perhaps oversight isn't a worse alternative.


Cheers

Greg Locock


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

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

 

[Ng2020]

Greg, not as far fetched as you might think: we do have tests for dynamic seat structure in civil aero (2x.562), which can include representative support structure.

 

[Sparweb]

@RB,
There's just as much GIGO in computational fluid dynamics as there is in finite element model structures.
My story:
About 6 years ago I undertook a contract for a US governmental agency intending to mount external sensors under the wings of a passenger jet. You might have heard me talk/write about this project previously.

I sub-contracted an engineering company to do the actual modeling because there were plenty of other project goals to satisfy, not just the CFD. After a few weeks of number crunching they came back "early" with their results and provided me with a gloriously colorful presentation. During the presentation my "spidey sense was tingling" but I'm not fast enough to catch all the flaws right away. I bought myself a bit of time by asking them to consider a few variants of the model they'd just presented. Then I immediately dusted off my old copy of XFOIL (or actually Martin Hepperly's JavaFoil) and cranked though a few 2D sims myself. That, plus a deep dive into charts in Hoerner and Abbott/VonDoenhoff's books, gave the the answers I was looking for. There's nothing like 80-year-old test data squashing a 21st century computer model, to remind you that we're designing planes for the real world, here.

The engineering company's model had shown the aircraft nearly in stall with only 10 degrees angle of attack. This was hard to figure out just based on blocky color cross-sections, until I could reproduce what they'd done. Even in 2D is was obvious. I figured out how to do sums to get a static balance and showed that the "model" airplane might generate enough lift force, but it had 5x more drag than the engines could react in thrust.

I was lucky to also have an intern with enough smarts to pore over their report and figure out what kind of crude turbulence model they'd been using. She came back in time with a different turbulence model to use instead. There were some remarks about meshing, too. I combined my (scathing) critique of their previous work with the new direction they should take. They re-did the model, and to keep them honest I specified a number of additional criteria to validate their results.

That was my trial-by-fire with CFD.
In case you're wondering, it was a little company you might have heard of known as "RAND corporation".