Practicing FEA Without a License

by Roopinder Tara, Tenlinks.com

CAD companies are eager to put analysis in the hands of designers. It's scaring me to death.

Yet Another CAD/FEA Demo
A bunch of us editors are sitting through yet another demo of how easy analysis is, being done on an automotive piston. I'm on pins and needles. Will we get it right?

I've done a few analyses of my own. I've made many mistakes doing so. Not just small mistakes, but whoppers. It is so easy to add an extra zero, or leave one out. It's also incredibly tempting to believe what the computer tells you.

Mesh? What Mesh?
A piston is subjected to a pressure load, restraints are applied, material properties checked. A mesh flashes on the screen. Oh, that was an accident. I was not supposed to see the mesh. The computer handles the mesh. 

"But doesn't the quality of the mesh affect the results?" asks one editor.

CAD companies are hiding much more than the mesh. Much of FEA now happens "under the hood." Why get your hands dirty, they seem to ask? In fact, I don't think I've heard any say "finite element analysis" for a few years; they all call it "simulation" -- and it's push-button easy.

We see the results, a stress contour plot. I note it is showing a von Mises stress, as von Mises accurately depicts failure in steel. The demo jock leans back in his chair -- job done!

High Stresses Occur Unexceptionably
But I can't figure out why the highest stresses are occurring in hot spots along the piston ring groove, in what appears to be an area free of stress concentrations. I ask if they might have been where restraints were applied. Incorrectly applied restraints can cause singularities. There is a bit of a scramble to explain, but my questions are not welcome. The demo is over and I'm clearly running the risk of being a serious buzzkill.

I am forgetting my press hat; I'm thinking of how I agreed to protect the public when I got my PE [professional engineer] license. So I persist in my righteous, seemingly futile crusade against practicing FEA without a license, a crusade I have pursed for years -- to the irritation of many a host.

"Have you heard of 'throwing a piston?" I ask, referring to what in my limited knowledge of engine mechanics seems to be a common failure mode. This is where an over revving engine will generate enough forces to shear the pin holding the piston to the arm, causing it to break through the engine block and, in the best case, dent the hood; in the worst case, cause injury and death. Maybe you should be analyzing for that, I feel like shouting!

Yeah, it's just a demo. With a "real" analysis, a real analyst would have had a chance to check the results, right? Call over a grizzled old veteran who has done a million of these analyses, maybe on a calculator, on paper, or even a slide rule. Some who know instinctively know how a part will fail and why. He'd have sent you back to your cubicle with your tail between your legs when you can't defend your analysis and crucified you when you stupidly applied the wrong failure theory. You'd have licked your wounds. Years later, you thank God the old coot was there and that those early analyses never left the room. A plane did not fall out of the sky. 

Those grizzled veterans are still out there, right? Is someone  checking all these fancy stress contour plots?

(Reprinted with permission of CAD Insider.)

Comments

[ralphg]

Rendering is not as dangerous to life and limb as is FEA, but I have wondered about the accuracy of renderings. For instance, just today a CAD vendor updated their rendered with these corrections:

-- more light in interior scenes that include some daylighting
-- the elimination of shadow artifacts in corners

[Steve Johnson]

I think Roopinder is selling the profession short when he says he was forgetting his press hat. I'd say it is an important part of the press's job to ask awkward questions, and to follow them up if answers are not forthcoming.

If press (and bloggers, for that matter) just sit back and unquestioningly swallow whatever we are shown and told at demos, what is the point of being there? Other than the free trip, obviously? Why not just stay at home, reprint the press release, pretend it's news and have done with it?

I'd still like to know why there is a stress hot spot in that piston ring groove. Roopinder, please ask again; it's a very relevant question to anybody who might be interested in buying the software. Is it an error in the design, in the use of the software, or in the software itself?

If you don't get an answer (a real one, I mean), then we need to know about that, too. That way, potential buyers can draw their own conclusions from any truth-avoidance and act accordingly, for example by practicing purchase-avoidance.

[Ken Elliott]

I took training on a rather specialized FEA package, using a stand-alone version and a CAD-plug-in version. The results for a well-understood device seemed correct for the stand-alone version, but the plug-in was way off the mark. Off enough to kill the product line - if we trusted the results - which we didn't.

I mention this to share the observation that even products from the same vendor (said to share the same solver code) can give vastly different answers.