WEIBULL AND BOOTSTRAP BASED PROBABILISTIC FATIGUE LIFE MODELING OF STAINLESS STEEL UNDER PWR COOLANT WATER ENVIRONMENT CONDITION

摘要

Current fatigue life design curve in ASME Boiler &Pressure Vessel Code Section III is based on the best fittingcurve of in-air data. Fatigue life of reactor componentsunder PWR water condition are estimated based on thisASME in-air design curve and a scalar/deterministicenvironmental correction factor. In the present paper, wediscuss a probabilistic fatigue life modeling approachdirectly using the fatigue data in the PWR water conditionsrather than using the deterministic life correction factorand the in-air ASME design curve. The probabilisticapproach has some advantages. For example, it canexplicitly quantify the uncertainty associated with scatterin strain versus life data points. The scatter in data pointsare often resulted from use of different grade and heat ofmaterial used for test specimens, test temperature, etc. Useof a deterministic environmental correction factor may notalone estimate the life of reactor components accuratelydue to different material grades, heat, surface finish andvarying operating temperature and loading rates(associated stress/strain rates) of actual componentcompared to the material grades, heat, surface finish, testtemperature and stress/strain control rates of the testspecimens (which associated data points were used forconstructing the deterministic design curve and theenvironmental correction factor). Hence, probabilisticestimation of strain versus life curve and environmentalcorrection factor might capture the bulk effect ofuncertainty associated with material grades, heat,environmental conditions, etc. of test specimens and actualcomponents, without explicitly factoring out the individualeffects which sometime not only highly complex but alsoimpractical (e.g. while factoring out the effect of surfacefinish in test specimen versus the actual component). As anillustrative example, we estimate the probabilistic fatiguelife model described by a modified Weibull probabilitydistribution function and using the stainless-steel fatiguedata under PWR water conditions. The maximumlikelihood estimation method is used for the parameterestimation, and the bootstrapping method is used for theuncertainty evaluation of the model. Finally, the resultingprobabilistic model is compared to the ASME design curveapplied by the environmental life correction factorapproach described in NUREG/CR-6909 Rev. 1. Theapproach described in this paper borrows some of theconcepts from upcoming technologies such as from datamining and artificial intelligence technologies to tryimproving the predictive capability of material/componentreliability.