When do you use FEM 3

[Kwan]

FEM analysis is quite expensive compared to hand analysis (for a simple problem). Just wondering how people in this forum decide when to use fem over hand analysis.

I have used fem on very large structure such as a 777 fuselage. However, this is a loads model not a stress model (i call it a stick model, no solids). Or perhaps to get a stress concentration on a very complex fitting (where i'll use a fine mesh model).

Anyway, just curious where most people draw their line.

 

[GBor]

If it can be hand calc'd without significant simplifying assumptions, hand calc it. If you have to overly simplify the problem, FEM it for the additional details. If you are trying to optimize on weight, I think FEM is helpful (I tend not to put in so many safety factors). If you are trying to figure out why something failed early, FEM it.

Bottom line, every case is different. I do FEM for a living, so I hand calc to make sure I am still using the software properly, but people come to me for FEM's. If it can be hand calc'd, I usually tell them so and suggest they save their money.

 

[JakeLiv]

as long as the analysis is static, and can be simplified, there is no real need for FEM.
In the real world, eventually you can into normal modes, frequency response and transient analysis. Those things are done only by computer FEM analysis, since any hand calculation is on a VERY basic level.

 

[EdR]

Actually you should take Gbor's comment a bit farther...Every time you do a FEM analysis you should do some sort of hand calc to check that the FEM model is not producing garbage....It may be as simple as a check that the total load is being applied to the model, the fundamental frequency (SDOF) is reasonable, approx. stresses are in the expected range, inelasticity occurs where expected, etc......and if something doesn't smell right that there is a reasonable explanation. Acceptance of FEM results without some sort of checks, regardless of how rough they may be, is just asking for big trouble.....

Ed.R.

 

[Kwan]

Thanks for the comments. I truely believe each fem requires a verification process. I like to check my model to test results and hand calcs (especially big airframe models). That's the best way to smell out bad models.

 

[KENAT]

One way to decide is if the contract specifically asks for it.

Had a very simple design I'd done a few basic calcs on. My chief stress engineer did a full analysis, again using hand type calcs (he may have put them in mathcad, can't remember).

However because of the way the contract was we had to do a FEM analysis. Took weeks, cost thousands and if I recall correctly didn't tell us anything we didn't already know.

 

[jte]

I tend to agree with the above responses: Many analyses are more appropriate as computer aided (i.e. Mathcad) hand-calc's. I could write it all out by hand but I've learned to like doing it once in Mathcad and then modifying as necessary later.

As alluded to in the post above regarding contractual obligations, sometimes the client wants an FEA. Most of my clients are "internal" to the company. There are times when its simply good politics to run a quick hand calc to determine the answer, then make some pretty FEA "cartoons" to make the senior level managers who most likely do not have a stress analysis background happy. They just feel better when they see a pretty picture. One of my favorite quotes: "Red is bad, right?"

jt

 

[cbrn]

Hi,
it depends... Please design a new ball valve for hydroelectric machinery and let me know if you can optimize this in less than 3 years using ONLY hand-calcs (or even computer-aided explicit-analytical maths)...

Regards

 

[meshparts]

Very good answer, cbrn!

I think, FEM can bring you there, where hand calcs can not.

Regards,
Alex

 

[GBor]

I think the bottom line is that none of us should be "purists". Each application should be looked at with engineering judgement...anyone remember that terminology? It's when we, as engineers, actually had to THINK! Computers are wonderful T..O..O..L..S, which, as every craftsman knows, are only as good as the craftsman behind them. I've had engineers with master's degrees hand be perfectly good-looking GARBAGE because they didn't apply reasonable logic to the computer result...

OK, down off the high horse!

I wouldn't hand calc too many assemblies unless they are fairly simple, but I do trust some of the event simulation software out there today. As engineers, we have to make certain we apply the correct theory to whatever hand calc we may do...thick plates do not behave the way Roark's says plates behave because Roark's concentrates on thin plate theory making some assumptions about the shear deformation. Timoshenko makes some assumptions about shear in beams that standard beam theory does not...which is appropriate for your application? On...and...on...and...on, etc. etc.

Garland

Garland E. Borowski, PE
Borowski Engineering & Analytical Services, Inc.
Lower Alabama SolidWorks Users Group

 

[EvanGr]

The whole idea behind the use of FEM is to model complex structures more accuretely than with hand calculations.So you have to balance the triangle of cost-time-quality. If the client is willing to pay for and bare the extra time cost then FEM should be adopted. But the client will for sure expect good quality of modelling and analysis.

Regards,
vangelis

 

[krishp]

Agaian, everybody used to have their own way of calculation. When FEM was not there people were depending on hand claculations only.They built so many famous structures which are still in existing. So I would say it is possible to calculate the things by hand which can be done by FEA...and when to use FEA and when to use hand calc...is only depend on the requirement of how fast do u need a solution and how much you are able to spend on the solution. As cbrn said, there exists difficult calculations but however its all depend on the requirement when to use FEM and when to hand calc..

cheers,
Tobias

 

[cbrn]

Hi,
just for the anecdote: when I ironically asked if anybody could calculate a ball valve for hydroelectric power-plant "from scratch" in less than 3 years using only hand-calcs, it was because in the company I work for, this has been done exactly in 1938.
When I had to re-calculate (cross-check) this in 2005 with the latest techniques available (i.e. FEM), I found that the deformation results determined by hand (using finite differences, I presume) corresponded to my FE results within a 2% error, globally... So, the results were "the same" but the time needed was 3 years in 1938 and two weeks in 2005... Should I repeat the same calculations right now, it would take me one week due to the improvements in hardware...

Regards

 

[prost]

Since you have reduced the analysis time from 3 years to 2 weeks, does that mean you get paid 78 times as much (3*52 divided by 2 weeks)?

Also, since your analysis is faster, have the problems become that much more complex, thereby negating speed advantage? I know that as computers speed up, what we analyzed has become more complicated mostly through our own choices to _not_ simplify structures, thereby negating any improvement in CPU speed or memory. In many cases (say airplanes), what was considered the analysis detail accepted say in the 1930s is no longer accepted by the manufacturer or the regulating authorities. I suppose the more complex analyses gives us better designs, but can't say for certain. There are an awful lot of airplanes designed 'back of the envelope' in the 50s that are still flying today.

Bravo to you for trying to repeat an analysis, though. I've never been able to find financial support for such endeavours. So many analysis methods are based on legacy analysis or experiments (for instance, Peterson's) that might be outdated, but no one seems to be willing to check the assumptions and try to improve the results.

 

[cbrn]

Hi,
Prost, of course the cost for doing an analysis with the "same" precision as "yesterday" has dramatically dropped. But moreover, it's the "time-to-market" which has dropped even more. So, in my example, no-one would wait 1 year TODAY for a design that nobody would have claimed to have in shorter than 3 years YESTERDAY. That's why we can't afford TODAY to calculate "by hand" a very complex structure, although this may be possible. In fact, at least in my example, this HAS been possible indeed. And, probably, you were as safe in a new Boeing 747 as you are now in a new Boeing 777, but the improvements are elsewhere: aerodynamic efficiency, fuel savings, time-to-market, etc...
Regards

 

[prost]

Surely you'd have to agree that the complexity of these analyses has jumped up substantially? Even if you are analyzing the exact same part, you are doing 100s more parameter variations (loads, geometry, etc.) to define the space of possible solutions. Just the quantity of information alone increases the complexity of your analyses; never mind the complexity of the structure you are analyzing and the myriad of assumptions you have made to make the problem tractable.

Computers allow you to do many more analyses in shorter time; are the designs really improved? As the design margins go to zero, how's your numerical reliability? That is, how reliable do you think your computations are? I'm not talking about numerical convergence; I am talking about your squeezing of safety factors (or margins of safety) by assuming that because you can model increasingly complex structures, your numbers are better. The reliability of your numbers can be controlled by the quality of your inputs--say you are doing a plasticity analysis, how good do you think that plasticity model really is, and how much do you trust the model inputs such as yield stress?

Sometimes maybe we just need faster ways to do the same hand calcs, for instance in Excel or Matlab.

 

[GregLocock]

I think the classic example where FEM wins over hand calcs is in optimisation. For instance, the casting for a suspension arm has some well defined loads and stiffness requirements. It is very easy to set the optimiser up to go away and eat away all the redundant material in the design while I sleep.

Alternatively in a spaceframe chassis design, it is very easy to create models that will run through all the permutations of tube diameter for each member to pick the best. (Well OK it can't, the number of combinations is too large, but there are ways of getting round that).

I'm not saying a great analysis by hand could not have reached those designs, but it is nice to be able to take a step back from the grunt level work and spend more time on the two important steps that can't be automated: 1) how good are our data and assumptions and goals (2) how sensible is the result


Cheers

Greg Locock

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

 

[FEAHappy]

It always good to use fem if you have he software and are quick with it.

If the geometry is complex; beyond a cube or a cylinder

FEM forces you to state the problem and think it through.
Many engineers that I have seen, who do not do FEM, do not do free body diagrams or many pages of hand written substructured sketches. Even a good engineer, without a free body diagram can miss something.