Deterministic and probabilistic modelling of brittle fracture mechanisms in ferritic steels

Abstract

The paper shows how an understanding of the micro-mechanisms of fracture, quantified by deterministic models, can be used to provide a physical basis for probabilistic assessments, particularly when the results of a limited data set have to be extrapolated to low failure probabilities. The variability in results may reflect the effect of random experimental errors in measuring what is essentially a single-valued parameter; it may reflect spatial heterogeneity in the source material; or it may reflect both factors simultaneously. Results obtained for high-quality, pedigree steels and model micro-structures tend to typify homogeneous behaviour, with variability dominated by random experimental errors. More general sampling from structural steels can, however, introduce a further degree of variability, resulting from spatial heterogeneity. This affects the validity of extrapolation of fracture toughness values to low failure probabilities. © 2006 The Author. Journal compilation © 2006 Blackwell Publishing Ltd.