analytical fatigue analysis

[api503]

i'am using proE/mechanica to conduct static analysis and then fatigue analysis of a complex structure.

i'm a little naive about the procedure for a fatigue analysis analytically.

i conducted the static analysis using Al7075 and SAE 1045 and also the fatigue analysis using the static results. I got the results in terms of log life in pro/E fatigue advisor.

"my problem is that since i have no experimental data to validate these results i wanted to validate it using analytical results". The procedure for the analytical method is way too confusing and i'm having a hard time trying to figure out what stresses to be used and the factors that affect the results from fatigue advisor in pro/E.

my hurdles:
1. which stress from the static analysis(max, min principal or von mises) do i use to calculate the number of cycles to failure analytically.
2. the S-N or the E-N approach to calculate the number of cycles to failure.
3. how do i incorporate or compare successfully the analytical and the fatigue advisor results.
4. is it possible or am i aiming too high.

hope someone can help me out.

 

[rb1957]

you can used canned fatigue solutions, that'll work directly from the FEM, like ncode. but where's the fun (or learning) in that.

find online fatigue analysis articles, there are plenty. try MIT opencourseware.

assuming your static analysis is for a typical (ie fatigue) loading, then you need to figure out the stress cycle ... the max and min (maybe min = 0). S/N (or E/N) depends on material and depends on the stress you've pulled from the FEM (is it reasonably the peak stress at the tip of the stress concentration or is it a typical stress (so you'll need to apply a Kt in the fatigue calc).

typical fatigue calc would sum (n/N) ... miner's rule ... if you have a single fatigue load it's easy (this calc combines different fatigue loads and cycles, like say 20% load for 100,000 cycles, 50% load for 10 cycles, ...). I don't know Pro/E fatigue advisor but i imagine that it's telling you that the life of the part is X cycles.

Quando Omni Flunkus Moritati

 

[GregLocock]

The success of a given approach depends on how closely your load history resembles that for which the technique was debeloped. In automotive chassis durability Miner's rule is hard to apply and doesn't seem to correlate very well. If your construction is not homogenous steel again you may find it doesn't fail in the same way.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[realtime2]

An addition to the previous posts, which I agree with, would be :

1.: max, min principal or von mises
Probably best to use von Mises. As rb1957 suggested you need to worry about the
Minimum Load too. Probably you calculated for some Max load? The question to also ask
is what is the stress at Minimum Load? If your load inputs are simple then just get
Max Load --> Smax
Min Load --> Smin
using the ratio of Max and Min loads.

2. S-N or Strain-N
Probably your computed stresses are Elastic? If you want to use S-N
for very short life (under 10000 cycles to fail) it won't work well for
any material that is not very hard. If you are targeting long life it will
probably work ok. You will need to find a Constant Amplitude stress life curve
for your material. Check on-line. If your hot spot is near a weld you can use
the IIW curves.

Strain-N
Given your elastic stress calculations for Smax and Smin: A plasticity correction
needs to be made. Typically people (and commercial programs) use the Neuber method
to correct elastic computations for material plasticity. The objective is to compute the
local (hot spot) max stress, and min stress, AND the local max strain and min strain
for your design load cycle using the material's cyclic stress-strain curve.
Again stuff is available on-line to do this.

Miner's rule doesn't work all that well when one has a mix of large and small load cycles,
but presumably you have only one Smax and Smin cycle set?

3. Typically people use a safety factor of 4 on predicted life, but it depends on
experience. Material scatter alone is about a factor of two, assuming wrought Alum or Steel.

4. If its a critical problem you should find a colleague to guide you.

 

[bxtsafe]

Hi api503!

I don't want to dishearten you, but FE-Based Fatigue Analysis, which is the fatigue analysis using the results of stress (or strain) of a FE analysis, is a really hard issue! I've just got my Master of Science degree in this subject and it took me 3 years of intense study.

Regarding your questions:

1) I recommend you to use the von Mises criterion as the Equivalent Stress. Calculate the EQUIVALENT ALTERNATE STRESS using the equation of the slide 31 of this presentation:

https://www.efatigue.com/training/Chapter_10.pdf

. For calculate the EQUIVALENT MEAN STRESS, use the equation 10.11 of the slide 34 of the same presentation. After that, use the SWT criterion (equation 7(a) of this article =>

http://www.fatigue.org/minutes/Spring-2005/DowlingBrasilPaper.pdf

) to calculate the EQUIVALENT ALTERNATE STRESS WITH R=-1 ( = Sa,eq ). Then use this equivalent stress in the Basquin Equation.

2) Basically, if the predicted life of the most critical point in your component is bigger thatn 5000 cicles, than use the S-N approach. Otherwise, use the EPS-N approach.

3) I would recommend you to validate the PRO-E response using a simple model. Try use a unnotched specimen subjected to axial ciclic load with R=-1. Of course, using R=-1 you will not need to use the SWT CRITERION. It's important to use an unnotched specimen ( like this =>

http://www.google.com.br/imgres?esp...hBwwAw&iact=rc&dur=925&page=1&start=0&ndsp=34

) .... because the we don't know if PRO-E calculate automatically the notch factor. Use a big radius in the specimen in such a way that the stress concentration factor will be Kt=~1 and at the same time we will assure that the maximum stress will be at the center of the specimen. Consider that all the MODIFYING FACTOR are equal to 1, to facilitate the correlation.

Good luck!!