Fatigue analysis of discretized structures with interval uncertainties under stationary random excitation via surrogate model

The fatigue analysis of linear discretized structures with uncertain axial stiffnesses modeled as interval variables subjected to stationary multi-correlated Gaussian stochastic excitation is addressed. The key idea is to estimate the interval expected fatigue life by interval extension of an empirical spectral approach proposed by Benasciutti and Tovo [1], called 0.75 α-method. The range of the interval expected fatigue life may be significantly overestimated by the classical interval analysis due to the dependency phenomenon which is particularly insidious for stress-related quantities. To limit the dangerous effects of the dependency phenomenon a novel sensitivity-based procedure relying on the combination of the Improved Interval Analysis via Extra Unitary Interval [2] and the Interval Rational Series Expansion [3] is proposed. This procedure allows one to detect the combinations of the bounds of the interval axial stiffnesses which yield the lower bound and upper bound of the interval expected fatigue life for the stress process at critical points of bar connections.