In this paper, a Bayesian procedure is proposed to quantify the modeling un
certainty, including the uncertainty in mechanical and statistical model se
lection and the uncertainty in distribution parameters. The procedure is de
veloped first using a simple example and then is applied to a fatigue relia
bility problem, with the combination of two competing crack growth models a
nd considering the uncertainty in the statistical distribution parameters f
or each model. This Bayesian failure probability analysis can be incorporat
ed with information from nondestructive inspections performed on the struct
ure to derive more realistic reliability estimates. The procedure for updat
ing the mechanical model, probabilistic model, distribution parameter stati
stics and reliability is illustrated for the fatigue reliability problem. (
C) 2000 Elsevier Science Ltd. All rights reserved.