Test time reduction - Ask any vibration guy in the world

[brianpaul]

I have a question. I'll keep it brief to not lead the response. However I would like to learn some of the background to the response.

Here is the question (please fill in the blank):

If you ask any vibration guy in the world the underlying premise of test time reduction through amplitude exaggeration he (or she) will respond with _______?

And why?

 

[electricpete]

I must not be a vib guy because I don’t know what you’re talking about. I can guess you’re referring to a concept similar to accelerated life testing, as applies to fatigue damage. The simplest model of fatigue is S*N. If you double the stress S amplitude you can achieve same fatigue in half the N cycles. If endurance limit is present, this approach remains conservative since fatigue effects which would not be present at the lower stress level can appear at the higher stress level.


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[brianpaul]

electricpete,

Sorry for my lack of clarity. The test time refered to is shaker table test time.

Curiously I think your rule of thumb for doubling the stress S amplitude to half the N cycles is interesting. As a mechanical I would have applied a different ratio.

 

[electricpete]

You're right - increasing stress by factor of 2 would decrease fatigue life by far far more than factor of 2. Thanks for the correction.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[IRstuff]

We've gotten one endurance vibe specification with:

(newPSD^4)*newTime = (oldPSD^4)*oldTime

which is consistent with MIL-STD-810F Method 514.5 Annex B.

Given that this is a "sanctioned" approach, I would expect most engineers in this field would

TTFN

FAQ731-376

 

[electricpete]

Given that this is a "sanctioned" approach, I would expect most engineers in this field would

would what?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[brianpaul]

IRstuff,

Thanks for your response. What you've listed is the method for determining the new spectrum to acheive the new test time.

What I am looking for is people to identify "the underlying premise of test time reduction through amplitude exaggeration"?

 

[GregLocock]

Not a fatigue guy. But...if you look at the curve of high cycle fatigue failure stress vs cycles for steel then roughly n.s^5 is the relationship of interest.

In practice this means that if you have a time history you can identify the small number of events that are actually doing the damage and discard the rest. Note that this is not INCREASING the stress, it is just selecting the important events, so you can run the system at damaging levels in less rig time than in real life. Typically out of a 3 hour representative durability cycle I can get 60% of the damage in 30 seconds. For convenience sake I actually use s^4.n

The other option, which you are discussing, of overstressing the part, relies on knowing that that s n curve is accurate for your system, which if it is not a steel bar undergoing bending fatigue may well not be the case.

On the other hand if you have something more nearly akin to a strength problem then overstressing may be an acceptable way of generating representative failures.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[IRstuff]

The "method for detemining the new spectrum to achieve the new test time" is the underlying premise.

TTFN

FAQ731-376

 

[GregLocock]

Actually the answer to the OP's first post is "are you paying for my rig time by the hour?"

He didn't actually say he was doing that weird PSD testing thing.



Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[brianpaul]

Okay, I will request forgiveness for the vagueness of my question.

The answer I was told to expect was "Miner's Rule". We do spectral exageration as IRstuff indicated via

(newPSD^4)*newTime = (oldPSD^4)*oldTime

or someother exponent. However how is that arrived at. GregLocock headed towards the derivation with n.s^5 or n.s^4.

My problem is that I don't understand how Miners Rule:

Sigma(ni/Ni) = D

plays into the derivation of the spectral exageration from n.s^4.

 

[GregLocock]

Ok, that exponent comes from eyeballed analysis of fatigue tests on steel specimens. As such it really varies according to the materials that are failing.

Miners rule gives you a way of adding up the contributions of different levels of stress for different numbers of cycles. Any textbook or even wiki will give you details. I'd add that it is not much used round here, the way we geenerate test loads seems out of kilter with your approach, probably becasue we have known operational loads. All we are trying to do is survive our defined physical durability circuit, the assumption is that if we do that then out in the real world the car will be fine. That's the theory. As such our tests are time domain, compressed by some arcane method to reduce the test time by eliminating the non damaging events.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.