Fatigue reliability analysis for the crack propagation compared with LRFD specification

Abstract

In this study, a fatigue reliability is assessed taking into account the uncertainties in load and resistance of a steel specimen, in which fatigue crack propagates. The results of the deterministic fatigue crack propagation has been compared with AASHTO LRFD specification. A response surface method (RSM) combined with an advanced first order second moment method has been applied in order to assess the probability of the remaining life of the specimen under cyclic load as a function of crack length. For composing limit state functions, the stress ranges, stress intensity factor and the remaining life was selected as input random variables for the developed fatigue-reliability analysis program. The choice of Bayesian Belief Nets (BBNs) as a tool for assessing uncertainties in resistant component of a limit state function allows an extended probabilistic assessment for the resultant fatigue life of a target specimen, in terms of resistant components of stress range, stress intensity factor, and material property. Additionally, fragility curve models are proposed to compare the probabilistic fatigue life as describing the probability of a structure to suffer a given damage level when it is subject to a given fatigue life. The proposed integrated stochastic models can significantly contribute to the design and maintenance of infra-structures, demonstrated in the numerical example in terms of damage index with the probability of exceedance the considered damage levels.