Influence of statistical and constraint loss size effects on cleavage fracture toughness in the transition—A single variable experiment and database

This paper, which is the first of two parts, describes a comprehensive experimental study to systematically measure constraint loss and statistical size effects on fracture toughness of a typical low alloy pressure vessel steel in the brittle-to-ductile transition. Both of these size effects lead to higher toughness in smaller specimens. Previous studies primarily involved specimens that differed in size, but with self-similar standard geometric configurations, making it difficult to separate constraint loss from statistical size effects. The focus here is on a single-variable database developed for a homogeneous plate section of the decommissioned Shoreham Reactor vessel steel, tested at the 3/4-plate thickness position. Tests were carried out at a common set of conditions and covered a full matrix of specimens with six thicknesses (B) ranging from 7.9 to 254 mm and four widths (W) ranging from 6.4 to 50.8 mm. The variations in B were designed to provide readily observable statistical sampling effects, while variations in W result in fracture at deformation conditions ranging from small scale yielding to those involving extensive loss of constraint. A detailed analysis of this database is presented in a companion paper, where physically based models are used to quantify the relative contributions of constraint loss and statistical size effects. The models also provide a general approach to adjusting measured toughness for both testing and applications. The B–W database and analysis also provides a comprehensive independent assessment of some key aspects of the master curve method in the ASTM E 1921 Standard discussed in detail elsewhere. In this paper, we present an empirical analysis that shows a high constraint toughness scaling with B that is reasonably consistent with ASTM E 1921, as well as effects of loss of constraint that occur at much lower deformation levels than reflected in the data censoring limit in ASTM E 1921.