Fatigue life bounds for randomly excited structures with interval parameters via sensitivity analysis

The present study addresses fatigue life assessment of linear finite element modeled structures with uncertain parameters subjected to stationary multi-correlated Gaussian stochastic excitation. Uncertainties are modeled as interval variables so that all response quantities, including the expected fatigue life, turn out to have an interval nature. The problem formulation relies on the interval extension of an empirical spectral approach, known as α0.75− method, which expresses the expected fatigue life in terms of four spectral moments of the critical stress process. The mean-value effect is taken into account by applying a suitable transformation of the Power Spectral Density function of the critical stress process. The bounds of the interval expected fatigue life are efficiently evaluated by applying a sensitivity-based procedure in conjunction with the so-called Improved Interval Analysis via Extra Unitary Interval. The presented approach requires only two stochastic analyses of the randomly excited structure for assigned values of the uncertain parameters. Such values are identified by sensitivity analysis as the ones which yield the lower bound and upper bound of the interval expected fatigue life. A steel telecommunication antenna mast under wind excitation is selected as case study.