Random Vibration Fatigue Cycles

[SlipperyPete]

If I run a stress analysis with a random vibration input spectrum, the output stresses will be in Grms (1-sigma), and therefore for a fatigue analysis I can also determine the 2-sigma and 3-sigma stresses. However, the thing I'm uncertain about is how to determine the number of cycles at each stress level. It seems unrealistic to base the number of cycles on the peak input frequency. Alternatively, I can calculate the resonant frequencies, but since this is effectively a flat random profile, all resonances will be excited simultaneously.
Any suggestions gratefully received.

 

[GregLocock]

That very much depends on the part, the industry, and accepted practice. Here's wat the automotive industry uses for road type inputs

https://en.wikipedia.org/wiki/Rainflow-counting_algorithm

Cheers

Greg Locock


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

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

 

[SlipperyPete]

Thanks for the response Greg. However, rainflow counting is used to reduce a complex set of periodic loadings to a simple set of stress reversals. I'm not sure if or how this can be applied to a random input spectrum.

 

[amanuensis]

You have to get first a time signal of your random vibration input spectrum.
Then you need an algorithm to count every time the signal passes through zero.
Divide the number, given by your counter, by 2 to get the number of cycles.

 

[GregLocock]

... this is, of course, nonsense, in the fatigue world.

Cheers

Greg Locock


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

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

 

[amanuensis]

Greg, you annoy me with your "nonsense".

Various methods of counting have been proposed in the literature, leading to different results and therefore (for some) to errors in the calculation of the fatigue lives.
They are peak count, range-restricted peak count, level-restricted peak count, mean-crossing peak count, range count, range-mean count, range-pair count, ordered overall range method, racetrack method, level crossing count, modified level crossing count, peak valley pair (PVP) counting, fatigue-meter count, rainflow count, national luchtvaart laboratorium (NRL) counting, and time spent at a given count level.

As you see, rainflow count is just one method among a lot of other methods...

 

[GregLocock]

So you are annoyed. Stop spouting irrelevant or misleading rubbish and I'll stop calling you on it.

Cheers

Greg Locock


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

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

 

[GregLocock]

Perhaps I'd better expand on what I meant - in my admittedly limited experience with the fatigue community, characterising a random signal merely by counting zero crossings is such a simple technique compared with what is usually done, and the absence of literature commenting on its usefulness and accuracy, leads me to believe that it is too drastic a simplification to be useful. It is, however, often part of what is done after the signal has been simplified in other ways, so as to be able to run Miner's rule or equivalent.

As a corollary to this, I strongly suspect I can synthesise a signal that is statistically equivalent to a given random signal, but that has a consistently greater number of zero crossings, and I know that when real time based fatigue loads are prepared for a rig test many zero crossings are eliminated.

Cheers

Greg Locock


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

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

 

[amanuensis]

You're not a gentleman.
When you write some lousy messages, I say to myself : " This guy is just an idiot" but I don't write it to the entire community.

 

[GregLocock]

If you don't like the way public forums work for professionals then don't participate. I'm here to learn.

Cheers

Greg Locock


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

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