SOME SIMPLE APPROACHES TO RELIABLE FATIGUE DAMAGE PREDICTION

The process of predicting safe fatigue lives for rotorcraft components is traditionally carried out by analysis with conservative assumption s made on the three stochastic areas of usage, loads and strength. Spe cifically, usage must be assumed severe enough to cover the worst oper ation over the intended lifespan of the fleet with respect to maneuver ing flight, gross weight, center of gravity, density altitude and rota tional speed. In addressing loads it is customary to gather representa tive samples for each regime in the usage spectrum, with actual fatigu e damage predictions made assuming the highest measured flight load (o r damage rate) in the sample is always present, Finally. to represent strength, a mean S-N curve is obtained from testing a small sample of full-scale laboratory (bench) specimens (typically 6) and reducing the working curve by three standard deviations (assumed to be mu - 3 sigm a) in order to assess the damage. Recently a lot of interest has been generated in the area of health and usage monitoring systems (HUMS). W ith regard to real-time fatigue damage assessment of dynamic system co mponents it is impractical, from the standpoint of weight, complexity and overall system reliability to monitor loads directly in the rotati ng system. This realization has lead to various techniques for monitor ing usage by ''regime recognition'' as well as methods which attempt t o derive actual rotating system flight loads using fixed system measur ements. In general all of these algorithms require vast computer resou rces to store and retrieve digitized time histories. This paper will i nvestigate the feasibility of two simpler approaches which utilize fix ed system statistics to predict fatigue damage. The two methods are: ( 1) Regression-based load synthesis with regime recognition coupled wit h a damage simulation algorithm; (2) Regression-based load synthesis w ithout regime recognition coupled with a damage simulation algorithm. The reliability of each method will be tested using randomly selected blind data sets. Both a main rotor (MR) and tail rotor (TR) dynamic co mponent will be examined in both steady state and transient flight reg imes. The flight loads data used in the study were acquired from fligh t testing of a USAF H-53.