Probabilistic assessments of fracture toughness in pressure-vessel steels

The paper describes microstructurally based models of brittle fracture and "fast shear" ductile fracture in steels and discusses the ways in which microstructure could affect the statistical distribution of fracture toughness values measured on sets of nominally identical testpieces. A distinction is drawn between "quasi-homogeneous'' material, in which it is argued that the distribution is Gaussian, and "heterogeneous" material, comprised of "tough" and "brittle" spatially-distributed "patches". Here, it is tempting to treat the data as if they belonged to a single distribution e.g. a Weibull. Such a treatment may be sufficient for interpolation, but can prove to be misleading if used to provide the basis for extrapolation to very low fracture probabilities, which is often required for safety assessments.