Estimation of Structural Component Loads in Helicopters: A Review ...

DSTO-TN-02396. ConclusionLarge variations in load were observed for the same helicopter type under the same flightcondition. The load variations were caused by variations in pilot technique, adversary type(in aerial combat), control sensitivity, mechanical assembly tolerances, wear and tear, grossweight, altitude, rotor speed, ambient aerostatic conditions, and aircraft response to name afew. Dynamic component loads are very dependent on the magnitude, rate, and phasing ofcontrol movements. Furthermore, analysing only peak loads within each manoeuvre (thesepeak loads often incur high fatigue-damage) exhibited larger variability than analysingcycle-counted data from either each manoeuvre or an entire flight. Using current fatigueestimation techniques, it is difficult to avoid over-penalising component replacement timeson most aircraft to protect against the worst possible usage for any given aircraft in thefleet. In fact, component replacement times might need to be reduced by as much as 75% ifload variation is taken into account. Thus, there are good reasons for taking into accountthe powerful effect of load variability on life in fatigue life expenditure calculations.In the frequency domain, the holometrics technique involves the development of a linearapproximation (or a calibration matrix) of the loads in dynamic components from the loadsin fixed components. The loads from fixed components are only included in the calibrationmatrix if they are sufficiently close to being orthogonal to the so-far-calculated calibrationmatrix. That is, the loads from fixed components are incorporated into the calibrationtechnique one at a time, starting with the most influential load, and sequentially adding thenext load that is closest to being orthogonal to the set of already collected loads. In theapplication of the holometric technique to an Apache helicopter, excellent identification ofdynamic system loads was achieved. The only exceptions to this excellent identificationresult were the identification of loads that were well below the endurance limit. Both aparallel vector reduction algorithm and a cost-function (which determined optimumcombination of inputs and hardware) provided an improvement relative to the originalholometrics technique. It is unclear why this apparently promising technique wasn’tfurther pursued. It appears that using parameters other than just loads proved moreproductive for other researchers.Several authors approximated the loads in dynamic components using regression of severalkey flight parameters. Most studies focused on high-G and high-load manoeuvres (whichconsume the majority of fatigue life). Results were improved by including terms in theapproximation that involved the coupling of parameters and filtering out low-G data (tocreate a high-G bias). Alternatively, different weightings of data allowed the model to bebiased in favour of higher loads. Measuring the swashplate servo position (which takesaccount of both pilot and stability inputs) instead of control stick position (which takesaccount of just the pilot input), and using aircraft accelerations instead of angular rates, alsoimproved results.One study developed a model using flight parameters and signal features (such as mean,maximum, and standard deviation). The main limitation of this work was the need to25