FE reports as reference and "Mesh safety factor"..

[SofieneFEA]

Hello All,

When doing a Finite Element study (of a steel component for example), suppose that we have a very good mesh and a very good model, I want to know if there are safety factors related to the finite element study to measure the difference of the results in comparison of the reality?

I would like to know also if there are complete finite element studies (reports) that have served in building a real part (that exist…not just a study for “fun”) that can be seen and serve as a reference for further studies?

Best Regards.

 

[gfbotha]

Not sure what you mean with your first question but FEA remains an approximation of reality - there will be an error of some value... There are uncertainties re. material input values, actual loading conditions, etc. Typical error for a typical linear static analysis might be 3 or 5% (but it really depends upon a lot of things).

2nd question: You might want to search for so-called benchmark problems or studies. NAFEMS published studies are a good reference.

 

[corus]

Reality is a strange beast. If you're referring to reality as actual measurements then they too can be subject to error and in some cases have to be interpreted to really get a real reality. There are no safety factors applied to FE models to compare with such reality. There may, however, be safety factos employed in designing a component for the 'unexpected' or worst possible case.

In industry where FE analyis is carried out you won't find out that a study has been done for 'fun', though I sometimes wonder.

corus

 

[SofieneFEA]

I totally agree with you Corus that in the industry the FE studies are not done just for “f”. I’m looking for FE studies done for industrial purposes to establish a trustful methodology.

The problem of benchmarks is that they are based (the result is compared to) on theoretical/analytical calculations.

So, if a mechanical part is designed to resist a certain load and to remain ender certain displacements (deformed shape) limits by the FE methods, it will really resist or not? In some cases, was it certain surprises (not resulting from incorrect modelling, meshing…)?

 

[GregLocock]

In the auto industry it had better perform like the FEA says. And that definitely includes static stiffness.

Cheers

Greg Locock

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

 

[corus]

FE models have different uses. Sometimes they're used to compare results due to proposed changes in a design, to look at a failure, or it may be to check a new design. The main objective is to have confidence in the answers you get. In the first two cases you can compare the results with measured values before you assess the changes to a design. In the latter case you can have no measurements so the only way to be confident of the answers is to compare against theoretical values at some part of the model. To have trust in the model you need one or both of these methods of validation/verification. NAFEMS gives good guidance as to best practice in using FE, but at the end of the day it's real validation of the model you want. There can be no trustful methodology as such as each model is unique. The only methodology I use is my common sense and whether the results make sense. If I don't spot an obvious error then I'm sure the non-FE user will spot it in 2 seconds.

corus