Scatter factor as per AC 23-13 A 1

[Jnnal]

Hi, as per AC 23-13 A, Eq.4 provides a scatter factor to account for the statistical variability inherent in the fatigue performance of built-up structures. As per the same, the standard deviation to be used for high strength steels like one used in landing gear are to be 25% and, the scatter factor worked out to be as high as 17. It looks very high. Generally, I see, aircraft industry uses 4 or 5 as a scatter factor for landing gear fatigue life estimation. Then, why is this discrepancy. in that case what number to be used for certification? is there any reference for the same. Thank you.

 

[rb1957]

I have rarely (rarely? well, just about never) liked 23-13, and this seems to be another dislike.

Typical safe life factor for landing gear is 3, for airplane structures 5 (sometimes 4).

Personally I consider fatigue calcs a very simplistic approach, and would rather rely on a crack growth from a 0.005" typical imperfection (with all the hand waving that goes along with DTA. For proof of fatgiue calcs rest in the very limited number of fatigue problems since we started using safe life principles.

But where did you get you standard deviation from ? It seems high to me.

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

 

[Jnnal]

Hi rb1957, It is given in the 23-13 itself, for various materials, it has been in the range of 14% (Aluminum) to 25% (High strength steel)

 

[rb1957]

I would not use it. Common practice with landing gears is safe life factor = 3 (sometimes 4, sometimes 2?).

I guess there is an AC for 571 that maybe defines safe life factors ?

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

 

[SWComposites]

AC 23-13A has (presumably) conservative std deviations, which are independent of alloy (well, its really not good how they defined the table as I'm sure the original data behind the table, from the 1970's IIRC, was probably only for a few materials); the FAA will presumably accept use of these values without questioning it. If you want better values, then you need to run a fatigue test program.

I don't think there is any other AC that defines scatter factors for materials. If MMPDS has fatigue data for your material you might be able to derive std deviations or scatter factors from it.

 

[Jnnal]

Thank you for your reply. However, AC 23-13 A says about " Standard Deviation (SD) of Log Fatigue Life for Built-up Structures Constructed of Common Structural Metals". How do get these SD. Is it from material test alone. Again, in S-N curve, we may have different SD at different stress levels. Is there anu particular reference is available. If we want to MMPDS, which stress level SD to we choose. is it the highest? Another problem is about batch variability. do we have to test different batch of materials like in case for A- Basis and B-basis.

 

[rb1957]

what sort of plane are we talking about ? part 25, part 23, UAV ? drone ??

I have never needed to refer to 43-13 in any fatigue analysis.
Typical safe life factors as above, use 10 if you can afford it or it you really think there are a lot of questionable assumptions (that are not clearly conservative, to any SME).

Safe life factors account for all the variability you speak of.

Use S/N from whatever reliable source, MMPDS or other.
Use Kt data for whatever reliable source (Petersen is typical, I use ESDU a lot)
Use stress data from your own structure.
Mix together and bake !

The fatigue loading is maybe the biggest area for thought. Wing or Fuselage ? Pressurised or not ??

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

 

[SWComposites]

There may be some fatigue data analysis method info in MMPDS (don’t recall off top of my head). There are some other references which I can look for when I get home from vacation in a week.

Yes, you probably need to test multiple material batches.

 

[rb1957]

yes there is some, I use the commonly used Al alloys (2024 and 7075) where there is a good library.

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

 

[Jnnal]

Hi, Thank you. Please share any references when you find some time.

 

[rb1957]

any references to what ?

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

 

[Jnnal]

Related to any literature on how to arrive at the SD or if a particular number for scatter factor is to be used, where can I find it? If we want to certify a product, we have to provide references. It is challenging to convince the authority that we may use 3 or 4 or any other number. Thank you

 

[Sparweb]

If we want to MMPDS, which stress level SD to we choose. is it the highest? Another problem is about batch variability. do we have to test different batch of materials like in case for A- Basis and B-basis.

Fatigue data is neither A-basis nor B-basis. It is average values only. That is the whole reason to use Scatter Factors - to account for the probability of failure based on the average and estimating variability.

In the MMPDS, don't just look at the charts, also look through the text in the chapter (presumably aluminum) and in later chapter 9 the description of the methodology.
AC 23-13 is only applicable to Part 23 aircraft. If you aren't working on an aircraft <19.000 pounds, then look somewhere else. If you are, then you can use the recommended scatter factors without challenge unless your structure has poor quality control.
AC 43.13-1B won't help, but I believe the reference to it by someone above could be a typo.
AC 25.571 is important to read even if you aren't working on a Transport Category aircraft.
Tom Swift's 1990 paper Repairs to Damage Tolerant Aircraft provides fatigue charts and if memory serves they may be corrected for scatter factor already. Read it and find out.
Jaap Schijve wrote the book Fatigue of Structures and Materials which discusses this in great detail.

 

[Jnnal]

Thank you

 

[rb1957]

give us just a couple of details, to frame our answers ..

Part 25, Part 23, UAV, drone ??

professional project, or school work ? (either is fine)

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

 

[LiftDivergence]

I think the best thing to help understand where these factors come from is rooted in statistics itself.

"Typical" stress-life or strain-life data curves are 50-50 (or average) curves. That is, they are at 50% reliability and 50% confidence interval.

The "scatter factor" that needs to be applied should be the ratio of 50-50 life to 99-95 life or whatever statistical basis is being mandated.

These factors will be highly sensitive to statistical characteristics such as sample size, standard error of the estimate, etc.

You can check this for yourself by selecting some stress-life data that has those type of metrics listed. Then find a table of k-factors for one sided tolerance intervals for a log-normal distribution and compute the 50-50 life and the 99-95 life. If there is appreciable variance in the data you can end up with a big factor.

This won't be exactly correct if the example life you check is HCF because with fatigue data, the variance actually increases with life. More appropriate would be to apply a weight function like in NACA CR-2586, or use a Weibull analysis.

But you can get the idea. Those "scatter" factors recommended by the FAA are "typical" or "blanket" numbers which are meant to apply to broad classes of material.

But I would caution anyone that they do not apply to everything. It is entirely possible that your source of data (including data from MMPDS) can have high enough standard error or poor sample size to invalidate those "typical" factors.

Also remember that this factor we're discussing will only account for variability of life with respect to stress level. It does not inherently account for other things like stress ratio, type of loading, or surface condition. In my experience, "scatter factor" generally refers to a total product of all reduction factors that account for any differences between the test data configuration and the real structure.

Finally, while 14 CFR 25.571 does in fact say:

"Inspection thresholds for the following types of structure must be established based on crack growth analyses and/or tests, assuming the structure contains an initial flaw of the maximum probable size that could exist as a result of manufacturing or service-induced damage"

But you also have to consider which ACO you are working under and if they have any specific guidance. For example, SACO provides guidance for establishing an inspection THRESHOLD which considers statistically based stress-life or strain-life as one datapoint. So depending on your situation, these calculations will be required.

Keep em' Flying
//Fight Corrosion!

 

[rb1957]

Wow ! all I can is ... wow. LPS for LD.

that report seems to talk to fatigue analysis methodologies, not much on safe life factors ? (which I think the OP is interested in)

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

 

[Jnnal]

Thank you all for this lively discussion. My question is related to PART 23 class aircraft landing gear. I am a starter in this area and my understanding/Queries as of now is
1. Fatigue life and damage tolerance life, both have to be established for the structure
2. Landing gear is designed as a safe life structure. (Only fatigue life)
3. Scatter factor (SF) has to be applied on the un-factored life to certify a component. (Scatter accounts for everything to the level of structure, like variability material, process, assembly, size, any coating, etc.)
4. In that connection, During the analysis and testing, we may have to show that SF*Life, the structure will be sustained, to prove the component to authorities for certification.
5. From the above discussion, The SF used in the industry appears to be 3 -5 or one has to use the formula in AC23-13A
6. Is there any solid reference or literature to back any particular number? if not
7. Is there any method (experimental) to arrive at our own SF?

Thank you

 

[rb1957]

I think there is some logic behind SF=5 for aircraft structure. I think it was derived after the Comet accident investigation (from the 50s) ... something like 1.5^4 where 1.5 is the sum of various variation factors. There may be ways to rationalise it, but generally I find these "specious" ... rationalising an assumption with different assumptions and a lot of impressive math. There may be discussion in practical fatigue analysis courses. It may be these are generally accepted factors. There may be something in 23.57x, maybe an AC defining acceptable factors.

But you're not working in isolation ... ask someone where you are (the guy who gave you the job to do ?)

SF = 3 is typical for landing gear, I've understood the lower factor 'cause the weight penalty was considered, that the static design cases are already very conservative,, and the consequences of failure are less than a wing/fuselage failing (at 30,000').

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

 

[Ng2020]

The scatter factors quoted in ac23-13 are associated with a level of reliability and confidence acceptable to the regulator for compliance with 23.572, the AC touches on this in its discussion.

If part 23 landing gear - why are you doing safe life?