FEA gives an approximate solution? Why?? 2

[9865395996]

I have learned, heard,experienced that finite element analysis gives only an approximate solution .

Just i need to know that what are all the factors that leads to this approximation??

How can we improve the efficiency of FEA to the highest?

 

[BlasMolero]

Hello!,
In fact, FEA is an approximate method of simulation using the computer, but the good part is that everything is in your hand to make your results to be accurate or not. It depends of many factors: accuracy in the material properties used, if loadings & boundary conditions prescribed in your FE model are correct, if meshing approach used is reasonable or not based in the finite elements used in your model, also if the mesh density is coarsed or refined, if the solution type used is reasonable or not, etc.. Please note everything in real life is nonlinear, dynamic & transient, but is up to you to make your assumptions, in your hands is the admissible error of your solution.

In summary, yes, FEA is an approximate method, but is THE method, the best we have today, if your learn to use it in a professional way you will see that is great, impressing ...

Best regards,
Blas.

~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director

IBERISA
48011 BILBAO (SPAIN)
WEB:

http://www.iberisa.com

Blog de FEMAP & NX Nastran:

http://iberisa.wordpress.com/

 

[IRstuff]

I hate to burst your bubble, but EVERYTHING in analysis is an APPROXIMATION. You assume that all your bolts are identically loaded and strained, you assume that your materials are homogeneous and yield at the values specified, etc.

The fact of the matter is that no EXACT solution can be had, even for the abstraction of reality that is your "model." FEA is the closest thing to being a solution that you can possibly have, for all problems beyond the trivial ones.

TTFN
faq731-376

 

[IceBreakerSours]

.. factors that leads to this approximation??

Approximations occur at several levels:

a/ mathematical model of the physical problem
b/ discretization
c/ applied loading, boundary conditions, interactions
d/ material model (isotropic/anisotropic .. linear elastic/hyperelastic .. )
e/ analysis (static/dynamic/viscoelastic,..)
f/ perhaps, the least worrisome (in overwhelming majority of cases) numerical approximation

How can we improve the efficiency of FEA to the highest?

I assume you are a beginner, and if so, I would grab a decent book, Bathe's for example, watch his online video lectures on linear finite element analysis first - available on YouTube, solve a bunch of simple problems for which analytical solutions are known.

Best.

 

[Twoballcane]

I agree with IRstuff. To add, FEA is just an interface tool between management and analysis. You should be doing hand calculations to understand what you are analyzing and then use FEA to confirm and show pretty pictures to management and the uninitiated. I suggest that you truly understand the math and physics that goes with the analysis. FEA is garbage in garbage out. If you are unfamiliar with the math and theories of what you are analyzing, please (PLEASE) hire a person who does (Structural/Mechanical Engineer). Also, hand calculations and FEA will either get you 80% there or total failure. Analysis altogether is just your best guess. Actual testing of the device is your final and accurate answer.

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

 

[inline6]

firstly what is the "real" solution? IMO it is impossible to know. Physical testing to verify FE models also has its own problems with regard to accuracy.........

 

[rb1957]

sorry in advance, but ...

"truth ? you can't handle the truth ..."

i'd start with a basic text book on FEA; very quickly you'll see the approximations being made and solutions to minimise the error.

 

[PicoStruc]

Just note :

(1) Quadrilateral elements, other 2D/3D elements are approximation of the problem and the result depend of the meshing quality and refinement.

(2) For strucural engineer, the 2D/3D beam element is exact at node location (displacement and end forces). Exact means it give the exact same output than by hand calculation of the bernouilli beam solution.

 

[jhardy1]

PicoStruc said:

"For strucural engineer, the 2D/3D beam element is exact at node location (displacement and end forces). Exact means it give the exact same output than by hand calculation of the bernouilli beam solution. "

Which may be true, but of course, Bernoulli Beam theory is only an approximation to the real world!

http://julianh72.blogspot.com

 

[Qshake]

picostruc - don't be quick to categorize....SEs are far more advanced than just 2D and 3D beams.....We've been known to use plates and shells and knives and forks and tension/compression elements, brick elements, all kinds of spring elements and on and on...

Regards,
Qshake

Eng-Tips Forums:Real Solutions for Real Problems Really Quick.

 

[izax1]

Can anyone define the exact solution? Even reality is very seldom repeatable, and thus the exact solution is hard to define. As engineers, we are responsible for defining the solution to a problem that will fulfill the requirements. I think asking what is the exact solution is the wrong question. Rather: Will this solve the problem? I have never encountered a problem that can not be soved with approximate solutions. But then we can always discuss: How close do we have to be? Hove reliable are the answers? Have I used adequate factors of safety?

 

[Twoballcane]

Analysis in general should be the search of minimums and maximums i.e. the minimum plate thickness to carry the maximum load and then add margin by factor of safety. Thus you’re looking for ranges of possibilities rather than exact solutions. You’re trying to make the best mathematical guess you can before you build and go to test. By doing the analysis and knowing the ranges, you will know the weak points of the design before going to test. With this knowledge, you can come up with a better test procedure to monitor the week points. The advantage is you can correlate the analysis and test data and for the next round you will get closer to guessing the measured physical phenomenon (deflection, strain, frequency, temperature…etc.). However, theories like stress, fatigue, PE and KE energy, heat rate…etc are still mathematical guesses with measured test data.

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

 

[PicoStruc]

Postjhardy1... Of course but generally when we are taking about the approximation/Exactitude of FEM... It's in comparison to best known model... not reality

Qshake... For sure we use other elements are used to by structural engineeers... see slab & shear wall design, stair design, custom built girder, etc.... What I meant is that beam are generally more used by structural engineer that any other kind of enginner.

 

[Nonlinearer]

When FEA is world wide an acceptable method of analysis then I don't think we can conclude something by discussing it why it is approximate.

IR

 

[Twoballcane]

Ummm where is FEA not accepted?

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

 

[IceBreakerSours]

Nonlinearer:

One of the take home messages, for a beginner or even an intermediate level FEA person, is that: If you are not aware of the assumptions, you are playing with a time bomb.

 

[PicoStruc]

IceBreakerSours... It worse than that !

One word... Ignorance !

> the bomb is already exploded... The problem is that you aren't aware of his existance yet !

Read that :

http://opinionator.blogs.nytimes.com/2010/06/20/the-anosognosics-dilemma-1/

 

[IceBreakerSours]

I stand corrected!

 

[SWComposites]

"Ummm where is FEA not accepted? "
> often FE results are not accepted by the FAA until the model is correlated to test data

 

[inline6]

"often FE results are not accepted by the FAA until the model is correlated to test data "

rightly so, there are two many morons out there doing FEA..........it's scary.