Spectra – to count or not to count - one mystery sorted

by Tom Deiters 15. January 2013 18:12

This subject does not strictly apply to the nuts and bolts of AFGROW but really applies to all damage accumulation schemes. It is a subject that occurs frequently enough that a discussion should be presented.

Questions continue to arise with respect to spectra. One is in respect to cycle counting and another one with respect to the use of exceedances (i.e., the use of TWIST or MiniTWIST or FAA Technology Center Loads Reports or MIL SPECs). Both questions have to do with a misunderstanding in spectrum development using exceedance data – not to be confused with raw flight test data which must be cycle counted for crack growth analyses.

For this discussion, I focus on those spectrums which will be or have been developed or attempted to be developed using exceedance data. For instance, I assume that an exceedance curve or curves for an aircraft of similar weight, propulsion arrangement, and wing arrangement, to the one needed was obtained. I also assume that everyone reading this post is familiar with what steps must be taken in reconstructing a maximum minimum load or stress spectra from exceedance data. TWIST reconstructs a spectrum based on probability distributions. But I recognize that there can be a large variation of techniques used and I’ll get around this by saying that however it is done, it is probably fine as long as it suits the intended purposes. Of course, the chosen method will no doubt account for ground and flight exceedances. The bottom line is I assume that by whatever process chosen, a crack growth input spectrum file has been developed that contains a few to as many as thousands of lines, each line containing one maximum minimum load or stress cycle … now let’s make some life predictions.

But wait before doing so … Mr. Hanalyst argues that cycle counting must be conducted but, Mr. Danalyst argues against all cycle counting. However, they both are right. What?

The crux of their argument is in the misunderstanding and assumption of what the crack growth input spectrum file contains.

The basis of all the rhetoric is not “knowing” or “remembering” or “recognizing” what exceedance data is. Exceedance data is just reduced data. It is reduced from flight measurements or reduced analytical data. It is reduced in that all the really important data has been eliminated, i.e., the mean load level shifts, all the segment to segment transitions, all order of load appearances, have been eliminated. By and large exceedance data represent just numbers and magnitudes of incremental excursions about a zero mean. Therefore, a proper spectrum based on exceedance data must be reconstructed to include those things which were eliminated. This continues to be the reason for all the heated debates.

So the basis of the argument for and against cycle counting is simply preconception. The following short list is offered as a job aid.

1. If an analyst has developed or reconstituted a flight by flight or block by block maximum minimum cycle spectra AND HAS ADDED the correct transition cycles (or GAGs) and has ordered the cyclic occurrences within the spectra, he DOES NOT NEED TO CYCLE COUNT. He can proceed directly with life predictions; he can carefully use load interaction options and make analytical life predictions.

2. If an analyst has developed or reconstituted a flight by flight or block by block maximum minimum cycle spectra AND HAS NOT ADDED transition cycles (or GAG) and has not ordered the cyclic occurrences within the spectrum, he NEEDS TO CYCLE COUNT. After the spectrum has been cycle counted he can proceed with life predictions. However, he should not carte blanche select load interaction options, because the cycle counted spectra has lost all appearance to a reality, i.e., flight by flight or cycle by cycle spectrum, in other words, the delaying action on subsequent cyclical damage accumulation would be suspect if not wrong. If he is accomplished and can somehow order the input spectrum so that the cycle counted spectrums maximum occurs when it should within the pass then load interaction can be used with caution.

Item 1 is true for all aircraft. Item 2 is strictly true for transport aircraft. The order of appearance of loads has not historically been seen to greatly influence fighter life predictions; in fact load interaction can typically be ignored on fighters, which leads to item 3.

3. If an analyst has developed or reconstituted a flight by flight or block by block maximum minimum cycle spectra for a FIGHTER aircraft AND HAS NOT ADDED transition cycles (or GAG) and has not ordered the cyclic occurrences within the spectra, he NEEDS TO CYCLE COUNT. After the spectrum has been cycle counted he can proceed with life predictions using load interaction or not.

Summary, if off the shelf exceedance data is used,

  1. Transition cycles have to been reintroduced in order to account for mean stress affects. Both ground to air as well as segment to segment transitions.
  2. The order in which the events occur must be accounted for. The reconstruction of mission profile segments should be as real as possible, for instance, taxi out, takeoff, climb, cruise, descent, landing impact, rollout, taxi back.
  3. If pressure is significant it must be superimposed on the applicable portions in the spectrum.
  4. If these steps are taken, cycle counting is not needed and load interaction can be correctly applied. If not THEN cycle counting should be conducted. The applicability of load interaction has to be considered carefully on a case by case basis.

Cheers.

Tags: ,


Spectra

Powered by BlogEngine.NET 2.0.0.36
Theme by Mads Kristensen