Probabilistic methods for forecasting development of fatigue damages of thin-walled aircraft structures, which suffer fatigue failures resulted from engine jet noise, pressure fluctuations in turbulent boundary layer or other high-level wide-band loads in acoustic frequency range, are discussed. These methods are based on capabilities of Markov networks. Development of fatigue damages is represented by continuous time, discrete state Markov processes with piece-wise constant interstate transition rates. These rates are used as free parameters to be identified during network training. The method of chi-square minimum is used to identify unknown network parameters, with observed and expected histograms of damage distributions being compared at the given time points. The technique in question is based on the accumulated observations only, without using physical theories of crack propagation. It gives new facilities for estimation of destruction processes in the aircraft structures under acoustic loads since do not use accurate load characteristics that are usually unknown and may be essentially changed at different flight regimes.
展开▼
