In general what is the accuracy of the results (eg. stress and displacement) produced by FEA softwares when the boundary conditions, material models, loading, mesh size are correct. At this time, I am only interested to know on a static loading case for a structural model in elastic region. I have done several structural analysis and i need to have some good correlation work to show the accuracy to non-FEA people to get convinced. Thanks.
Eng-Tips is the largest engineering community on the Internet
Intelligent Work Forums for Engineering Professionals
FEA results and correlation 5
- Thread starter Sripri
- Start date
For a simple, static case (cantilever beam) compared to, say, a hand calculation from Roark's formulas, you can be within 1%.
For more complicated geometry (something worth modeling and analyzing), you can be within 5%.
For more complicated analyses (non-linear), 10% is good.
Garland E. Borowski, PE
Borowski Engineering & Analytical Services, Inc.
Lower Alabama SolidWorks Users Group
For more complicated geometry (something worth modeling and analyzing), you can be within 5%.
For more complicated analyses (non-linear), 10% is good.
Garland E. Borowski, PE
Borowski Engineering & Analytical Services, Inc.
Lower Alabama SolidWorks Users Group
First of all, when speaking about FEM accuracy, a reference solution (either analytical or experimental) should be considered.
Most of the FEM codes provide excellent solution accuracy for linear elastic materials as compared to the analytical solution of the same problem.
As the material response gets more non-linear, obtaining an anlytical solution becomes impossible and assessing the accuracy of the FEM solution gets more complicated. The most objective case is to compare the numerical solution to experimental results. However, in many practical situations, it is quite difficult to obtain accurate experimental results. Therefore,in many cases, the validity of any reference solution may remain debatable.
The decent commercial FEM software have manuals including examples and verification problems showing the accuracy of the FE solution.
Most of the FEM codes provide excellent solution accuracy for linear elastic materials as compared to the analytical solution of the same problem.
As the material response gets more non-linear, obtaining an anlytical solution becomes impossible and assessing the accuracy of the FEM solution gets more complicated. The most objective case is to compare the numerical solution to experimental results. However, in many practical situations, it is quite difficult to obtain accurate experimental results. Therefore,in many cases, the validity of any reference solution may remain debatable.
The decent commercial FEM software have manuals including examples and verification problems showing the accuracy of the FE solution.
GregLocock
Automotive
As an example, I FEAd a ladder chassis for a car and compared it with a couple of real world results. The maximum error in the significant deflections measured along the ladder beams was around 15% worst case, and the predicted overall stiffness at the loading point was much better than that. We then added a stiffening brace and checked that we got the same % improvement.
That was beam elements, although I had previosuly modelled the critical joints as shells to get a good estimate of their stiffness.
The reason I used beam elements is that I then have an optimiser that tries to lighten the structure whilst obeying other constraints. This typically takes several thousand runs altogether, although most are discards. I haven't seen a free optimiser that will handle box like sections in a satisfactory fashion, that is why I used beam elements.
For modal analysis I'd expect to see major modes of a glazed car body within 10%, frequency wise, or better.
Cheers
Greg Locock
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.
That was beam elements, although I had previosuly modelled the critical joints as shells to get a good estimate of their stiffness.
The reason I used beam elements is that I then have an optimiser that tries to lighten the structure whilst obeying other constraints. This typically takes several thousand runs altogether, although most are discards. I haven't seen a free optimiser that will handle box like sections in a satisfactory fashion, that is why I used beam elements.
For modal analysis I'd expect to see major modes of a glazed car body within 10%, frequency wise, or better.
Cheers
Greg Locock
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.
Hi,
in addition to the interesting example from GregLocock, I can say that, any time we performed experimental measurements on FEAded components, the calculated deformations were always well within 10% error, most of the time the relative error is below 5%.
We are dealing with hydro-power machinery, so the parts never cope with material non-linearities (though we have several other types of non-linearities, such as contacts and non-holonomic restraints).
Regards
in addition to the interesting example from GregLocock, I can say that, any time we performed experimental measurements on FEAded components, the calculated deformations were always well within 10% error, most of the time the relative error is below 5%.
We are dealing with hydro-power machinery, so the parts never cope with material non-linearities (though we have several other types of non-linearities, such as contacts and non-holonomic restraints).
Regards
Hello,
In your case (linear static response of an elastic structure), the accuracy of the results obtained by a FE analysis depends mainly
of the data (boundary conditions, loading, material properties).
That means most of it you have listed as 'correct'.
If the data are 'correct', you can refine the mesh to obtain the required accuracy.
So I think your question is badly put.
If you have to convince non-FEA people, perhaps you could compare the accuracy of the FEM to other methods (analytical, numerical and experimental methods).
And there are analytical solutions for some non-linear problems.
Regards,
Torpen
In your case (linear static response of an elastic structure), the accuracy of the results obtained by a FE analysis depends mainly
of the data (boundary conditions, loading, material properties).
That means most of it you have listed as 'correct'.
If the data are 'correct', you can refine the mesh to obtain the required accuracy.
So I think your question is badly put.
If you have to convince non-FEA people, perhaps you could compare the accuracy of the FEM to other methods (analytical, numerical and experimental methods).
And there are analytical solutions for some non-linear problems.
Regards,
Torpen
The one thing most of the replies have forgotten so far is that it really depends on the analyst.....Most of the guys who replied are good analyst's and know how to get good results.....(mostly through experience and making most of the errors that can be made)....Someone who doesn't have the corresponding background and experience will most likely get garbage results....even for simple linear static problems....
Ed.R.
Ed.R.
- Thread starter
- #9
Thanks for the all the good answers. I was not having my internet for a few days. Explaining a little more on my earlier question, I am looking for stress accuracy for a complex part, which is still under elastic state when loaded. It is not a question of results not correlating, rather, how far the FE results can get close to actual when compared, if appropriate refined mesh, boundary conditon, etc are used. Thanks.
-Define "actual"!
A material characterized as "elastic" can be
linear elastic, non-linear elastic, described using a small or large strains formulations. Also, the strain and stress measures used in FEA can be different (e.g. Cauchy stress, Kirchhoff stress, 1st or 2nd Piola-Kirchhoff stress).
If you are interested in linear-elastic materials, analyzed using small displacements and small strain formulation (i.e. classical elasticity theory) for materials which are not incompressible, then , with appropriate modeling (meshing, boundary conditions, loading) you can get a very accurate solution.
A material characterized as "elastic" can be
linear elastic, non-linear elastic, described using a small or large strains formulations. Also, the strain and stress measures used in FEA can be different (e.g. Cauchy stress, Kirchhoff stress, 1st or 2nd Piola-Kirchhoff stress).
If you are interested in linear-elastic materials, analyzed using small displacements and small strain formulation (i.e. classical elasticity theory) for materials which are not incompressible, then , with appropriate modeling (meshing, boundary conditions, loading) you can get a very accurate solution.
Given a model made from a steel part how accurately do you really know Young's modulus and Poisson's Ratio??? If it's only within say < 5% then the best you could expect from any analysis would also be < 5%.....
The point is that there are a lot of variables in any analysis (some of them values we think we know) that can affect the results so it is not a question anyone can give an absolute answer to.....
Ed.R.
The point is that there are a lot of variables in any analysis (some of them values we think we know) that can affect the results so it is not a question anyone can give an absolute answer to.....
Ed.R.
EdR
You require Young's modulus and Poisson's ratio to convert strain guage readings into stress, but for a linear FEA Young's modulus has no effect on stress levels, only on deflections , and Poisson's ratio has only a marginal effect.
So I'm afraid your statement directly linking the accuracy of the material properties to the accuracy of the analysis is not correct.
You require Young's modulus and Poisson's ratio to convert strain guage readings into stress, but for a linear FEA Young's modulus has no effect on stress levels, only on deflections , and Poisson's ratio has only a marginal effect.
So I'm afraid your statement directly linking the accuracy of the material properties to the accuracy of the analysis is not correct.
-
1
- #13
I don't think your reply is correct either johnhors. Stresses will depend on relative stiffness within a structure and that will be goverened by the Young's Modulus or Modulii. Stresses due to temperatures are also related to the young's modulus by E.alpha.T
If you're using FEA to calculate stresses in a beam then you're right, but then the stiffness is uniform throughout.
corus
If you're using FEA to calculate stresses in a beam then you're right, but then the stiffness is uniform throughout.
corus
Sripri,
A little more detail about what the material is, what the part shape is and what you are trying to get across to common people would help.
A deflection test is the quick and easy way to determine modeling accuracy. If the deflection is correct the stresses show will be correct within a few percent. The only other way to know is to place strain gauges on the surface and load the part. If there is much shape detail that will be difficult or impossible. Otherwise a cycle test is the only other way to actually know the accuracy.
A little more detail about what the material is, what the part shape is and what you are trying to get across to common people would help.
A deflection test is the quick and easy way to determine modeling accuracy. If the deflection is correct the stresses show will be correct within a few percent. The only other way to know is to place strain gauges on the surface and load the part. If there is much shape detail that will be difficult or impossible. Otherwise a cycle test is the only other way to actually know the accuracy.
johnhors:
For any determinate structure the stress is independent of the material, however, for an indeterminate structure the stress is dependent on the material....Take any indeterminate structure (say a beam) and you will find that the reactions are a function of Youngs modulus....Then since the strains are the partial derivative of the deflections the strains are also a function of the deflections i.e. the moduli.....Of course the stresses are then a functions of the material properties and the strains so my original point is indeed correct for indeterminate structures....Guess I never thouoght about anyone using FEA to do determinate structures except for check problems (and school problems)
Ed.R.
For any determinate structure the stress is independent of the material, however, for an indeterminate structure the stress is dependent on the material....Take any indeterminate structure (say a beam) and you will find that the reactions are a function of Youngs modulus....Then since the strains are the partial derivative of the deflections the strains are also a function of the deflections i.e. the moduli.....Of course the stresses are then a functions of the material properties and the strains so my original point is indeed correct for indeterminate structures....Guess I never thouoght about anyone using FEA to do determinate structures except for check problems (and school problems)
Ed.R.
Follow-up to previous post:
After thinking for a minute even determinate structures are subject to variations in E.....The deflection for any determinate beam is a function of E....The only reason the stress is independent of E is that the value of E . eps is a constant for a beam (uniaxial stress).....and again the strains are the partial derivative of the deflections...etc....
Ed.R.
After thinking for a minute even determinate structures are subject to variations in E.....The deflection for any determinate beam is a function of E....The only reason the stress is independent of E is that the value of E . eps is a constant for a beam (uniaxial stress).....and again the strains are the partial derivative of the deflections...etc....
Ed.R.
-
2
- #18
Isn't stress always independent of 'E' in a linear problem? Given a notch or hole in tension, for instance, the stress state can be completely described by the applied far field stress and the geometry. Deflection is a function of the stiffness 'E' of course.
prost:
General 3-D Hooke's Law is typically written as a function of E & nu....even for linear problems (or the elastic part of non-linear problems)...In some special cases (like beams) it degenerates to the stress being independent of E .....
Ed.R.
General 3-D Hooke's Law is typically written as a function of E & nu....even for linear problems (or the elastic part of non-linear problems)...In some special cases (like beams) it degenerates to the stress being independent of E .....
Ed.R.
Prost, your quote :-
"Isn't stress always independent of 'E' in a linear problem? Given a notch or hole in tension, for instance, the stress state can be completely described by the applied far field stress and the geometry. Deflection is a function of the stiffness 'E' of course."
Is 100% correct !!!!
Ed.R.
In a LINEAR ELASTIC analysis, vary the Young's modulus (but keep Poisson's ratio the same), then for the same applied boundary conditions the resulting stresses never change, whilst the displacements are directly propotional to the value of Young's modulus. Whether the problem is determinate or indeterminate is of no consequence !!!
"Isn't stress always independent of 'E' in a linear problem? Given a notch or hole in tension, for instance, the stress state can be completely described by the applied far field stress and the geometry. Deflection is a function of the stiffness 'E' of course."
Is 100% correct !!!!
Ed.R.
In a LINEAR ELASTIC analysis, vary the Young's modulus (but keep Poisson's ratio the same), then for the same applied boundary conditions the resulting stresses never change, whilst the displacements are directly propotional to the value of Young's modulus. Whether the problem is determinate or indeterminate is of no consequence !!!
- Status
Similar threads
