Fatigue Data - 50% Probability Query

[morris9791]

Dear Experts,
I have a general query regarding approaches to take when there is a lack of sufficient data for fatigue analysis.

Is it the case that most fatigue data corresponds to average stress-life or strain-life curves fitted to a set of test data which relates to 50% survival probability?

If components are to be designed to 90% survival probability, how does one incorporate safety factors to account for this low 50% survival rate?

Currently my company uses a double figure to scale up the target number of cycles to be met. I don’t know where this figure comes from. Is there a more accurate approach to take in terms of assuming a reasonable safety factor far analysis?

Thanks

Ed

 

[unclesyd]

I would think you could find the answer here.

http://www.fatiguecalculator.com/

 

[tbuelna]

morris9791,

I can't answer your specific question, but you might want to take a look at how rolling element bearing fatigue lives are calculated statistically, and what impact various parameters might have on that life value. A 10% failure rate in your statistical sample would equate to an L10 life, and a 50% failure rate would be an L50 life.

In general based on my (limited) experience with calculating bearing life, I would suggest that the FoS=2.0 that you are using to compensate for the difference between a 90% reliability vs. a 50% reliability is probably a little low. As you get close to a 100% reliability, the fatigue life tends to get reduced very quickly. As an example, the fatigue life of a bearing at >99% reliability is only about 1/20th that of a bearing fatigue life at 90% reliability.

As I'm sure you're aware, calculating fatigue life is a very complex process, involving many factors. Without material fatigue data based on a statistically representative sample, your best bet is to be very conservative in your analysis.

Hope that helps.
Terry

 

[morris9791]

Terry and unclesyd,

Thanks for the useful post.
Terry, our FoS is not 2 but in the region between 10 and 15. Although I don’t know where the figure came from, would this be conservative and in line with what you have stated? For we generally do not have sufficient fatigue data that is statistically representative and hence why we use this ‘magical figure’!

Best Regards
Ed