Analysis for All, Part 1-- Should You Worry?

The fashion among CAD and FEA companies is to make analysis so simple that simpletons can use it. But as I fly home on a Boeing 757, I pray no such simpleton has analyzed the engine mount forgings or the (very thin-looking) landing gear struts. Similar thoughts occur to me when I drive across new bridges -- but I reassure myself that public works, at least, are signed off by degreed and certified engineers.

Am I the only one worried?

Finite element analysis was once the domain of an engineering department's special force -- the analysts. These over-educated geeks worked separately (often with their own big computers), dressed differently (white coats), spoke a different language (of "eigenvalues" and "discretization"). They seemed to worship a different god (Von Mises). Ordinary engineers would feed drawings through a slot in the door. Days or weeks later, an analyst would grace them with the results presented from a big stack of computer printout. Blah, blah, eigenvalue, blah, blah, sparse matrix, blah blah. The engineers would patiently wait for the ever important Factor of Safety, which would tell them whether it was back to the drawing board or on to production.

What has happened to the analysts? Perhaps they are in padded rooms and straitjackets, blubbering to no one in particular about planes falling out of skies as they imagine rooms of monkeys pecking at keyboards.

Comments

[Vince Adams]

While I used to say a number of the same things about part-time analysts, I've come to the conclusion that the upsides far outweigh the downsides. Keeping in mind that most companies still rely on prototyping as the final arbiter of product viability, what is the real danger of non-specialists taking a crack at up-front validation? I've expanded on this at the link below and would really like to see this conversation kept going!

-- Vince
http://designsmarter.typepad.com/vinceadams/2006/06/should_analysis.html

[Sugato Deb]

Validation with physical test is the key to managing the uncertainity associated with simulation, whether the source is the user (more likely) or the limitations of the software. Most organizations run their simulation efforts in open loop, validation puts it into a closed loop self correcting cycle. It also significantly improves the acceptance of simulation based decisions by other parts of the organization.

[Fat Guy]

But see look here. I have some very specific ideas about what kind of chair I want to sit in, and what kind of bed I want to sleep on. I am free to construct a chair or bed of my own design out of PVC or wood, etc. Nobody complains that I have no degree in chairmaking/bedmaking. What's wrong with me having access to a simple tool that can model my fat butt on a chair and give me some feedback about whether I am designing in the right direction?

[Vijay]

I agree with most of the comments I read here.

Especially the previous one which drives home a point about non-engineers inputting data. I have a few such clients.

A frequently asked question about our software, CAEPIPE, for pipe stress analysis, from newbies, is "How come your software does not tell me where the pipe is failing? Does it show that in the graphics?"

No. It does not and will not. Knowledge of engineering principles and software assumptions combined with sound judgment make for good analysis. When an engineer's expectations are met with results from the software then the engineer has a good understanding of the problem and the tool s/he is using.

This is a reason WHY we periodically hold seminars in Piping Design and Analysis; to dispel such myths and make the engineers more knowledgeable in the vast scope of piping analysis. Being able to model real world problems in light of current theory and software limitations is one of the goals of our seminars.

I just wish more dept. heads would see the value of such training.

[Billmce]

It has always been and will always be about process - not software. If your company's process lets bad analysis and bad product out the door.....well it doesn't matter a hill of beans what software you have or don't have or who used it.

They are tools - plain and simlpe. People that make a big deal out of them and say idiots will start making airplanes fall out of the sky haven't got a clue how airplanes are designed and built. Can mistakes be made sure but why do you think people test? Analysis is minimize the time you spend testing and gaining insight into how your product is responding.

If companies can't get a process together to properly exploit these tools I would say their longevity has uncertainties because their competitors will.

[Rob M]

Having been involved in demonstrating and selling FEA over the years I can tell you with certainty that while we believe that having FEA tools available for more people to use is valuable never once has anyone I know said that the tools can be used without understanding the fundamentals of FEA. In fact we still feel that there is a strong role for the analyst or specialist to play in reviewing the results. Having FEA tools on more desktops lets people refine their products more before conducting more detailed analyses and building prototypes.

Every time I introduce anyone to the concept of FEA on the desktop I tell a little story about a customer of a previous employer who used the tool without understanding the fundamentals and it lead to a very expensive conclusion. At our company we encourage FEA users to take refreshers in the fundamentals and to take advantage of the analysts out there to check results.

If one is prone to worry then don't worry about the fact that FEA is more affordable and more powerful and available to more people. In fact, be thankful that it is and that it means more things get studied. What you should worry about is companies, who are driven to cut corners and costs who arent investing in training their people to use the various design tools properly.

[Sean]

You should be worried. Anyone can get an answer out of an FEA program. However they answer can more easily be wrong than right. Getting a correct answer involved knowing how to properly set up your loads and support conditions.

I was always taught that to use FEA properly you need to know what the outcome will be. FEA simply proves that you are correct or gives you a more accurate answer than you calculated yourself. If they two do not jive then you need to examine your calcs and/or the model.

Too many software companies throw out analysis packages without regard as to how the users will be using them.

It scares me too.

Sean Dotson, PE
http://www.mcadforums.com

[ralphg]

As technology makes things more convenient, it allows lowers the barriers to entry -- which is good and bad.

Spreadsheets made it easy for me to optimize the mortgage for my home in 1991, something I would not have been able to do ten years earlier. (The spreadsheet model allowed me to find an innovative way of paying down the mortgage 2x more quickly -- one that even the bank's loan officer didn't know about.)

Same for desktop publishing, raster editors, and the like: I can create really nice name tags for my daughter's graduation reception. But neither example puts people's lives at risk.

Having amateurs think they can do FEA is a problem. While only experts should be involved, cost and time pressures may cause companies to relegate the work to the uninformed.

Perhaps there needs to be a barrier to entry for software that affects peoples lives. But where to errect it? After all, Myspace is being sued over an assult to one of its members.

[Miles Archer]

This has been a problem for at least two decades.

I have a buddy who works for a software company that makes stress analysis software. He used to get tech support calls from people who didn't know the basics.

Once he got a call from an engineer's secretary who was asked to input the design data to the analysis package. Naturally it was not just the case of typing in some numbers, but required understanding of what they meant.