Fatigue load spectrum 1

[jpaero]

Background: I am involved in a fatigue life extension project of an aircraft. The initial fatigue life of the aircraft was determined using a theoretical load spectrum of 10,000 FC. And now we want to evaluate the actual fatigue life consumed. So we are going to perform a few flights (in line with the A/C general usage) with instrumentation to measure the g's which would be then used to generate a test spectrum. This test spectrum would be compared to the original theoretical spectrum to determine the amount of life left in the structure.

Question : Because we are doing only a few flights (let's say 50 flights), we might not encounter some of the gust velocities that we might have encountered if we were to fly longer (let's say 100 or 200 flights). How do I account for events (like certain gust velocities) that are not encountered during the flight test?

 

[SWComposites]

You will need to determine a fatigue spectrum based on known or conservatively estimated load exceedances for the aircraft and its flight missions. Just using data from a few flights is not sufficient. And how are you going to determine the loads the aircraft actually experienced in its lifetime?

 

[Sparweb]

You are reproducing the methodology of many USAF and FAA test programs carried out on numerous aircraft. You could use the same methodology in the test and data reduction.

Fighter Jets:
Technical Memorandum AFFDL-TM-75-37-FBE Study of Exceedance Curve Spread ... by mission type ... flight data, 1975

Passenger Jets:
Final Report DOT/FAA/CT-94/21 Analysis of European Flight Loads Data, 1994

Small Aircraft:
Final Report DOT/AFS-120-73-2 Fatigue Evaluation of Wing and Associated Structure on Small Airplanes, 1973

There's a lot more like these, for a wide variety of aircraft, so finding one like yours is worth the look.

Assuming you may have already gotten this far, and found these materials, if they aren't giving you the details that you need, you could drill down through the referenced reports and underlying data. Some may be lost in the mists of time but not necessarily. If you went back a ways into the past, the NACA did tests such as these and they did publish their detailed methodology.

https://sti.nasa.gov/

Well it seems the site operators have managed to mess up the search engine on the NTRS site in the past year. Good luck, then...

 

[LiftDivergence]

What I would recommend is producing specific exceedance curves with range exceedance levels grouped into bins. If you are only accumulating instrument data from a few typical flights, what you will basically have data for is everything up to the typical or average once per flight max delta G level you can expect.

However, for fatigue and damage tolerance over the life of the aircraft (up to 10's of thousands of flights) you could statistically expect to encounter higher levels than that. Usually to produce a simplified spectrum for analysis, you would sum damage and get a weighted average over flights ranging from very light damage (typical flight, majority of weighted average) and also account for damage levels that might be expected only once out of 1000 flights.

With limited data you could either:
1. Perform statistical analysis of the data to try to understand the statistical distribution of the delta g occurences and then extrapolate
2. Try to come up with a conservative envelope of the gust exceedances per nautical mile based on the shape of the curves you have access to (either in the references mentioned above, or these type of documents: DOT/FAA/AR-98/28 Statistical Loads Data for Boeing 737-400 Aircraft in Commercial Operations

https://www.tc.faa.gov/its/worldpac/techrpt/ar98-28.pdf

There are a lot of other aspects to your comparison though. For example:
- What was the order or time history of the exceedances assumed in the original analysis which set the life? How are you going to order them? When do you assume the most damaging flights will occur? This is a debatable question.
- Was the original spectrum edited (by which I mean counted... rainflow, range-pair, etc.) Will you count it the same way? Seems like you'd want to for an apples-to-apples coparison.
- You are measuring the g's to determine the range exceedances, but how do you know what mean stresses were assumed in the original analysis, or that the way you operate the aircraft during data measurement is representative of this mean stress? Mean shift can have a big impact.
- How do you know what set the original life limit? Was the life limited by statistical fatigue (stress-life, S-n), or was if limited by LEFM and potential rogue macroscopic flaw assumptions? You could perform a fatigue life comparison based on the two spectra you are comparing, but that won't do much good if that's not the check which originally set the limit.
-How do you know what safe-life factors or scatter factors were used in the original analysis. Again, you need to use the same values to make the comparison relevant.

And that's just the tip of the iceberg...

Keep em' Flying
//Fight Corrosion!

 

[rb1957]

I would express caution.

1) First understand how the original flight was defined by the OEM. Was it conservative (max TOW) ? Was it Part 23 ?? What duration was assumed ? What limits the life ? (this could be an easy part to replace or it could be an airframe limit.) How much testing was done for cert ??

2) Why do you believe you can extend the life ? Can you point to something where the original analysis was conservative (weight, duration) ? I have recently done exactly this. The operator had TOW data (from the logbook) and I was able to show the OEM assumed MTOW (conservative) and a bit of "analysis magic" showed the airplane still had extensive life remaining.

3) You have in mind a new mission. Analyze this mission with ESDU 69023.

4) How will you approve your change ? presumably via STC ? This gives you some latitude to take the plane out of service, do your few test flights, and develop a new limitation. "a few test flights" doesn't define things well. I'd suggest a few test flights for some initial limitation, then loads monitoring (CG nz) to collet data for something like a year (a 1000 flights) to validate.

5) How will you analyze the airframe ?? or do you have OEM data to say this was the gust spectrum they applied (so something good to compare to) ??

6) is the plane pressurised ??

This is not "QED" and should have specialised engineering support, and not "advice from the internet". Where are you ??

another day in paradise, or is paradise one day closer ?

 

[crackman]

Hi jpaero

I concur with the recommendations from LiftDivergence. The process of developing a fatigue spectrum let alone a test spectrum is not a trivial effort particularly if one's aim is to extend the life of the airframe. I have seen several life extension test programs turn into life reduction programs. Obviously, access to complete detailed load histories are paramount and cannot be obtained from a single flight test. Typical load histories are presented in terms of 1000 hour blocks but generally require several thousand flights to accumulate. For instance, to obtain 1000 hours of taxi exceedances, it might require more than a couple thousand flights. It is possible however to use load histories from similar aircraft. However, great care must be taken in using them and ensuring that they are of like configuration, wing loading, gust response, etc.

One also needs to establish the mission types and utilization. This can come from many sources. The FAA DOT database contains over 400 aircraft and over 20 years of utilization data. You could also perform an operator survey as many have done. Additionally, depending on what parts of the airframe you plan on extending the life for, you need to identify all of the pertinent flight segments and conditions. Some cases such as braking and turning may only pertain to wing and gear while others such as rolls or approach conditions could involve flight controls. Either way, all flight conditions and pertinent fatigue loads and exceedances must be accounted for.

Assuming all of the above is accomplished, then there is a considerable effort in ensuring a proper flight by flight loading spectra unless of course you believe that there is a large amount of life left in the airframe in which case you can use something much more conservative. The flight by flight spectra then needs to be carefully truncated and clipped typically to be tailored to the test equipment requirements. Essentially, analytical spectra tend to be fairly large and generally cost and schedule dictate that the spectra be reduced but without losing any pertinent damage.

As you can see, its a fairly involved task and requires knowledge of several aspects. Boeing used to teach an in house course on fatigue loads and spectra. There is one other week long DTA course being taught recently which spends a couple full days teaching the methods for fatigue spectra development and reviews various full scale test spectra.

Anyways, good luck in your efforts.

 

[jpaero]

Thank you all!! The information in your responses is going to take me a while to digest!!!

 

[Ng2020]

To me it sounds like you're trying to make the argument that either:
1. The OEM's gust spectra was very conservative. There is a lot of gust data in the public domain for various aircraft types.
2. Actual historical usage has been substantially more benign than the OEM intended.
I don't think you can make the latter argument without really good usage records.

 

[jpaero]

@Ng2020: My question was related to methods of extrapolating flight measurements to cover for gusts that were not experienced during those flights. For example, if I do a single test flight to arrive at my 10,000 flight cycle usage spectrum, the test flight almost certainly will not experience all the kind of gust velocities expected over the 10yr usage. That also doesn't mean that I should do a 10yr test flight campaign. I was looking for the statistical jugglery available for that extrapolation of 'n' test flight data to 'N' (N=n.10,000).

 

[rb1957]

have a look at ESDU 69023

another day in paradise, or is paradise one day closer ?

 

[GregLocock]

You might have some luck plotting the rainflow data out on suitable axes and seeing if you can extrapolate to low occurrence high stress events. Trouble is the damage per cycle is of the order of s^5 typically, so small mistakes can have a big influence.


Cheers

Greg Locock


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

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

 

[crackman]

If you are trying to use a couple flight tests to extrapolate exceedance data for fatigue spectra, I can tell you, imho, that is a waste of time and will not provide any conclusive data.
There are upwards of 50 or more public references (NASA, DOT/FAA, RAE, DOD, ect) with existing large databases of dozens and dozens of different aircraft from cessna 172 up to 747. Your time would be better spent comparing your aircraft configuration, wing loading, etc to these aircraft and find one that is comparable and use that data. If the planned environment is a severe one, then instrument and collect a couple years or more of data and then re-evaluate at a later date.
good luck.

 

[verymadmac]

My question would be why remove the instrumentation after a few flights, there are numerous operators who continuously monitor aircraft cycles (Ag operators being one set that come to mind). Permanent instrumentation is likely cheaper than the cost of the engineering sweat required to make a 50 cycle sample work. If 5000 lb ag operators can wear the payload cost, a Business Jets should be able to (That's the only class that comes to mind for civilian aircraft with a 10,000 cycle limit).