Built-in Stress Effects on Fatigue Spectrum

[P2FModder]

Hi there,

I have a generic question. I have fatigue stresses for a particular aircraft which is derived by scaling "Reference Stresses" from FEM with spectrum factors (S_fat = S_ref x factor). Both of which were provided by the aircraft OEM. The scaled stresses are then used for fatigue and damage tolerance analysis.

If there is a need to account for built-in stresses due to manufacturing (which cant be avoided or removed by other means), should the built-in stress be scaled by the factor? Or should the built-in stress be added directly to the S_fat? Note that the built-in stress was also derived from the same FEM.
1) S_fat_new = (S_ref + S_builtin) x factor
2) S_fat_new = (S_ref x factor) + S_builtin

I have heard arguments for both sides. For point 1, the argument is either just because it is conservative (not considering penalty to inspection etc) or due to the fact that both stresses (S_ref and S_builtin) come from the same FEM and therefore should be added together prior to scaling it. As for point 2, the argument that I have heard is the built-in stress is not subjected to the cyclic loading and therefore should be excluded from being scaled by the factor.

What do you guys think? Thank you in advance.

 

[rb1957]

the difference is how much confidence do you have in your assessment of built-in stresses (something notoriously difficult to evaluate).

maybe you want a third (??!!) option ...

3) S_fat_new = (S_ref x fatigue_factor) + (S_builtin x builtin_factor) ??

All three are very similar ...

1) has builtin_factor = fatigue_factor
2) has builtin_factor = 1 (so you have high confidence in the builtin stress)
3) has separate factors for fatigue and builtin stresses (to reflect different uncertainties)

"but" builtin stress will "only" change the mean stress (not the alternating stress), so will affect R which is generally a weak variable.

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[SWComposites]

1) at what load level are the reference stresses? Limit?
2 exactly how are the built in stresses obtained from the FEM? Did this come from your analysis or from the OEM.
3) Calc of built in stresses is very difficult bordering on impossible, and are likely not deterministic but probabilistic, unless one is very conservative.

 

[Sparweb]

Presumably there is a S_fat_min and a S_fat_max otherwise there's no stress cycle.
And if there are pre-loads or residual stresses or whatever in the parts that you're accounting for, then S_fat_min isn't zero.

There are plenty of times when we assume that the structure "at rest" has "zero stress" and away we go with the assumption that stress ratio R=0. There are times when this isn't true.

 

[LiftDivergence]

If I'm understanding what you're concerns are, you are saying the manufacturing process leaves the parts with a residual stress at certain details which will affect the mean stress during the duty cycle.

If that is true, the "proper" way to approach F&DT with the spectrum factor approach is to include the accurate mean stress magnitude (from residual stresses) in the VA load spectrum during cycle counting and equivalent damage computation. That is, the stresses you're referring to should be "built-in" to the spectrum factor.

If they are not, due to OEM oversight on the manufacturing process or something, I suppose you could come up with another stress factor yourself. But to do that accurately, you'd need to compare accumulated spectral damage with and without the residual stresses present.

Simply scaling the manufacturing induced residual stress by the same spectrum factor and summing it with other stresses will not be remotely accurate, and I'm not sure it would be conservative.

You need to know a few things:
- Did the OEM actually compute the spectrum factor with the residual stresses you are concerned about in consideration?
- Are these residual stresses actually a concern? They may not be coincident with the critical fatigue details, right?
- What is the nature of these residual stresses? If they are compressive in nature, they may have a beneficial effect which would make this whole discussion simpler.

Note that the GAG stress cycle for the equivalent damage stress is typically set with an R = 0 fictitiously. That does not mean the mean stress in this equivalent stress cycle is accurate, it just means that's the magnitude required with an R=0 to get the single cycle equivalent damage.

In summary... if the spectrum factor you have was not computed with the residual stresses in mind, neither or your options 1 or 2 are correct. If the spectrum factor was computed with appropriate assumptions of mean stress, then your option 2 would be acceptable.

For example, a typical aircraft fuselage has sources of stress being 1. internal pressurization, and 2. inertial stresses such as gust and maneuver, which have associated spectra. It is true that often times these are separated out when computing a stress for damage tolerance, and a spectrum factor typically is only be applied to stress from inertial sources. BUT it is important to keep in mind... the spectrum factor would have been developed based on GAG damage and damage comparisons which account for mean stress contribution from the pressure stress.

Generally damage will be worst for a fully reversed stress condition (where mean stress = 0). But damage comparisons need to always be consistent. There are a lot of factors at play for fluctuating stress... mean stresses affect notched and unnotched components differently so damage ratios may not be the same for different assumptions of R.

It all comes down to what assumptions were made during the development of the spectrum factor... when the spectral damage was computed, was there an assumption that every exceedance had an attendant exceedance of equal magnitude and opposite sign, such that all damage is assumed based on R = -1? Or was the damage simply computed with an R based on the min an max values of each exceedance? We can't really answer your question without more detail on the factor itself.


Keep em' Flying
//Fight Corrosion!