In this paper a new model for the prediction of the Cumulative Distribution
Function (CDF) of fatigue life of structural elements during the crack pro
pagation stage is described. This problem is considered as a cumulative dam
age process following the probabilistic approach of Bogdanoff and Kozin [1]
(B-models). The initial and final crack lengths, the crack propagation ang
le, the material fracture and elastic parameters and the external loads may
be considered as random variables. In this initial approach, a linear appr
oximation of the random variable 'fatigue life' and a truncated uniform dis
tribution for the crack length variable are considered. Two corrections to
this model are discussed: a second-order approximation of the fatigue life
to compute its variance, and a modification of the Probability Density Dist
ribution (PDD) of the crack length, which is now derived from the truncated
uniform distributions of the initial and final crack lengths. Some example
s for mode I are compared to the ones obtained using a Monte Carlo scheme w
ith 400 000 samples, showing a good agreement and a much better performance
of the corrected version of the model, specially for big standard deviatio
ns. Copyright (C) 1999 John Wiley & Sons, Ltd.