Non linear analysis vs stress linearization

[tokitalk]

Got a question on the way to go to perform fatigue calculations. If some yielding is known to be taking place based on linear elastic analysis, do you linearize to deduce fatigue stress, or is say elastic plastic analysis a better option. What governs the choice of calculation method in these cases?

 

[rb1957]

actaully it's very easy ... you've run an elastic (linear) FEM, and got stress output. ok, some stresses are above yield, ok, we know that they're not real (you could do a NL FEA to find "real" stresses), but they are linear.

so what %age is your fatigue load of your static (ultimate?) load ?

 

[corus]

For pressure vessel codes, the elastic stress is used for fatigue assessment. This is probably due to the fact that they rely on hand calculations to determine the nominal stress. Strain-life curves use the elastic plastic results, but I've not seen this method in design standards.

 

[tokitalk]

If the stresses are not real, how do we know they are linear?

 

[rb1957]

the stresses are not real 'cause linear FEA doesn't account for yield and plaasticity, and so produces numbers greater than Ftu (which can't happen).

but they are linear, 'cause thats what the L in Linear FEA stands for. so if you want the results for 20% load you can divide the results for 100% load by 5. maybe this'll get you back to the real world. if you're still above fty, then you'll need a NL FEA to get the "real" stresses near your stress concentration.

or you could use a Kt factor ...

 

[izax1]

I think it is important to realize that as engineers we are never in "the real world." We are making models of what we think is reality, and our models will never be "real". So we use safety factors to compensate. This is particularly important if we use FE, which is farther away from reality than what we call "closed solutions" And closed solutions are only mathematically correct. I raise this because I have seen too any examples of engineers thinking that if they can use FE, the answers will be "right". That is of course not so. It is mainly a question of finding out: How wrong am I?
Fatigue analysis is largely based upon statistics, and I guess all agree that statistics is another limited model for the reality.
Well, I guess I got into the philosophical corner.