Full second-order approach for expected value and variance prediction of probabilistic fatigue crack growth life. (April 2020)

• Record Type: Journal Article
• Authors: Mallor, C.; Calvo, S.; Núñez, J.L.; Rodríguez-Barrachina, R.; Landaberea, A.
• Is Part Of: International journal of fatigue. Volume 133(2020)
• Publication Date: 2020-04
• Subjects: Probabilistic -- Fatigue crack growth -- Fatigue life prediction -- NASGRO -- Statistical moments; Materials -- Fatigue -- Periodicals; Materials -- Fatigue; Periodicals; 620.1122
• DOI: 10.1016/j.ijfatigue.2019.105454
• Languages: English
• ISSNs: 0142-1123

Abstract

Highlights: The extended FOSM to second-order gives efficient stochastic crack growth equations. Probabilistic NASGRO formulation to predict stochastic moments of lifetime is given. Stochastic analysis of axle's lifetime under random loading was performed as example. The method applicability considering loading spectrum was checked for expected value. The accuracy was examined showing good agreement with Monte Carlo simulation results. Abstract: The NASGRO crack growth equation for fatigue life estimation is ordinarily used in its deterministic standard form. This paper presents a new probabilistic formulation for fatigue crack propagation based on the NASGRO equation providing a stochastic approach for predicting statistical moments of fatigue lifetime. The methodology approximates the expected value and variance, of the output random variable fatigue lifetime ( N ). These moments are obtained from the approximation via the Taylor series up to the quadratic terms, full second-order, of the NASGRO equation with respect to the random input variables taken into account. The validity of the proposed method is verified by two numerical examples regarding the fatigue crack growth in a railway axle under two different random loading conditions. The first one applies the probabilistic formulations under a random bending loading. The second one takes into account a service loading spectrum acting on the axle subjected to a random load. Then, the statistical moments calculated are checked by comparison with Monte Carlo simulations. The probabilistic model developed enables an efficient estimation of statistical moments, providing accurate probabilistic results that can be used in design stages, reliability studies or in damage tolerance assessment.