what FEA is

[NikonF6]

I am noticing, increasingly, that FEA is totally mistaken in Mechanical Engineering. It becomes replacement for engineering calculation, replacement for stress engineers, anyone becomes an expert for stress analysis literary over night.... One take max stress from FEA result and directly compare it to UTS from webmat.com If FEA result is higher than such UTS it will break, ...

At area with notches, FEA result is not a real stress, it is theoretical stress containing "theoretical stress concentration". This stress concentration (actually it a form factor) has to be eliminated from FEA to get the real stress. To eliminate it one have to know the "form factor" and there are 3 trickeries to find out it, if not available on charts. This misleading can cost mass-production company millions... $$$$$$$$. But as I can see no one cares for it.

What is your experience?

 

[TGS4]

FEA is an excellent tool for getting very good looking, wrong answers.

Are your examples related to linear elastic analysis? Because there's more to FEA than linear elastic analysis.

I agree on engineers using it as a crutch for avoiding performing other engineering evaluations. Or for knowing fundamental strength of materials. Or even statics...

 

[IRstuff]

GIGO - garbage in, garbage out

This is nothing new, I heard almost exact same complaint about users of SPICE (Simulation Program with Integrated Circuit Emphasis) in 1977, who religiously believed in the output of the program, even if the output was meaningless.

TTFN
I can do absolutely anything. I'm an expert!
homework forum: //

http://www.engineering.com/AskForum/aff/32.aspx

faq731-376 forum1529

 

[csk62]

Not real stress at notches?

FEA is more than capable of modeling elastic plastic behaviour at stress concentrations.

 

[corus]

"At area with notches, FEA result is not a real stress, it is theoretical stress containing "theoretical stress concentration". No, it isn't. FEA doesn't have a list of stress concentrations within it that it applies to the stress. It is the real stress, or as real as it can be with the mesh you've chosen.

 

[NikonF6]

Corus
how you account for material sensitivity on stress concentration, residual stress from thermal treatment, material behavior in different directions, ...., Cast Iron does not have stress concentration in static load but FEA result shows it, ...

And on non-linear : one have to have FLOW-curve for the material (say some specific Steel) to be able to do it. Just using the program your non-linearity is along straight line (preprogramed). Ask guys that work in company dealing with forging, drawing,... what Flow-curves are. That guys dealing with non-linear all the time, even without computer.

Corus, Engineering is a little strange think.

 

[mmar2087]

There is a problem with overconfidence of FEA results from people using the softwares with low engineering knowledge and low knowledge of how FEA works. The cad-embedded softwares has increased this problem, not because they are bad, but because the education that the resellers give is quite crappy, focusing mainly on how to work the UI.

Singularities due to bad boundry condition or simplified geometry is unreal. Stress concentrations in many other cases are, as Corus mention "as real as it can be". One have to understand the limit with linear elastic theory before "comparing max stress with UTS". Using non-linear elasto plastic models can show you what happens with the stress concentrations after you reached the yield level. Bi-linear can be enough but of course using the S-S curve is even better and especially if you have this from tensile tests done on the actual material from the actual production method.

Stress concentration factors is something you use on hand calcs, not in FEA.

 

[NikonF6]

FEA program calculate stress concentration itself, when it exists, and incorporate it into stress shown on the screen. This cannot be avoided. This is used today to make stress concentrations charts in addition to already known and published charts. BUT this stress concentration is theoretical, not actual, and to get actual stress one have to divide the stress from FEA with that theoretical stress concentration. The further use of that "FEA" stress concentration is to calculate dynamic (fatigue) stress concentration. Stress to which that "NEW" FEA stress should be compared is not easy to figure out. It is calculated from UTS, Yield stress, ... which are available and ready to use. UTS, Yield stress, Shear limit,..., which we can find around (e.g matweb.com, books, ...), is for very carefully made material, not commercially available (Springs are different), and for standard sample. So that UTS, ..., has to be recalculated on real part. And so on...

The above procedure is known to "stress experts". However skipping it all and use just all as seen on the screen make the whole new army of stress experts emerging from (literary said) stupids.
FAE stress as seen on the screen, with a notched part, is with very limited use, and for linear and for non-linear analysis.

 

[ESPcomposites]

Finite stress concentrations are not "theoretical" in FEM. They are simply stress concentrations. If you are interested in fatigue analysis for metals, they are important. If you are interested in the ultimate strength of a ductile metal part, the stress concentration may (or may not) have a significant influence. This will depend on the volume of the material under the highly stressed region (and ductility of the material). This can be captured in FEA via elastic-plastic analysis (or disregarded with sufficient understanding of the material and specific problem). If the material is a composite laminate, then the stress concentration has a significant influence on the ultimate strength (but that is a more complex topic).

If the "stress concentration" is actually a crack tip or a location that would not show stress convergence (certain boundary conditions), this is actually a singularity and not a "stress concentration". In that case, you can either : improve your FEM to avoid this scenario, use fracture mechanics if this is a true singularity (and interested in determining its effect), or disregard the result if it is not in the region of interest.

In the end, FEA will give you exactly what you ask of it. But you need to understand the difference between a stress concentration and a singular stresses, how to interpret the results, what type of analysis to perform, and understand how the material reacts to the stress or singularity. The important thing is that you understand the classical methods before you perform the FEM (and not the other way around). Unfortunately, the "plug and play" FEM tools allow people to quickly generate stress plots, while a thorough understanding of the classical methods takes many years of engineering practice in an environment with knowledgeable people. So it is tempting for many people to go directly to the "plug and play" finite element analysis, without sufficient background understanding. However, that usually results in GIGO for anything but the simplest problems.

Brian

http://www.espcomposites.com

 

[NikonF6]

ESPcomposites
Finite stress concentrations are not "theoretical" in FEM. They are simply stress concentrations.

What is "simply stress concentration", how you define it?
Have you ever did manual stress analysis?
How you define "ductility of material" and how FEA take it in account?
How you verify FEA result?
Have you ever saw charts for stress concentration? and how you use it?

There is saying among Engineers: if you can't do it manually you can't do it with FEA neither...

 

[ESPcomposites]

- Stress concentrations are finite. FEM stress will converge to a solution upon mesh refinement. At a location of a stress singularity (as opposed to stress concentration), mesh refinement will not show convergence.
- Classical analysis (manual analysis) should nearly always be done and often compliments FEM. FEM is often used to enhance a classical solution as opposed to a standalone solution. Of course, there are exceptions to this (research versus production environments, etc)
- Ductility is the amount of plastic deformation a material exhibits. You can account for this with an elastic-plastic FEA (material nonlinearity). Alternatively, if you understand the problem/material/etc, you can sometimes disregard localized stress concentrations (as previously described). But you not should perform a nonlinear analysis without a thorough understanding of mechanics of materials and the concept of ductility.
- Verification of FEA can be done in many ways. That is a very broad question with a lot of answers. But in general, FEM is most appropriate when used within a domain that it has been used for with a past history of success. For example, for decades, aircraft companies have successfully used FEM's as loads models and to determine stress concentrations (and local stresses) for problems with unique geometry that can not be found in Peterson, etc. Alternatively, one can validate a FEM via test and then use FEA for derivatives from the original to minimize subsequent testing.
- Peterson's book is probably the most common source for stress concentrations. You use it as described in the previous post. Metal fatigue is a common application. In some cases, it can be used for ultimate strength (usually more appropriate for materials with low elongation or when a higher degree of accuracy is desired). For metals with large elongation (typical aircraft metals), stress concentrations are frequently disregarded for ultimate strength. However, this is usually slightly unconservative because the fracture strain is not infinite.

Brian

http://www.espcomposites.com

 

[NikonF6]

vague terms.
What is now finite stress concentration? Are you making up some new kind of stress concentration? Do you have any reference dealing with finite stress concentration.
There are two stress concentration: theoretical (or form factor) ALPHA, and fatigue stress concentration BETA. That is all.
Theoretical str. conc. is what you find e.g. Paterson's book, and FEA always incorporate it into stress. Try to calculate stress manually and compare it to FEA result. THE SAME stress you will get. But it is not to use in stress analysis. One have to make further work on FEA to get real stress.
Linear FEA deals only with Modulus of Elasticity and Poiso. ratio. This is by far not enough to define stress, e.g if material is with fibers or not, you will get the same stress. This is not the case in reality. FEA can't recognize Cast Iron lamellar or spheroidal structure. FEA IS O_N_L_Y INPUT INTO ENGINEERING CALCULATION, not output stress.