A systematic investigation of the effects of specimen size on the cleavage fracture toughness of a typical pressure vessel steel is reported. Size dependence arises both from: (i) statistical effects, related to the volume of highly stressed material near the crack tip, that scales with the crack front length (B) and (ii) constraint loss, primarily associated with the scale of plastic deformation compared to the un-cracked ligament dimension (b). Previously, it has been difficult to quantify the individual contributions of statistical versus constraint loss size effects. Thus, we developed a single variable database for a plate section from the Shoreham pressure vessel using a full matrix of three point bend specimens, with B from 8 to 254 mm and b from 3.2 to 25.4 mm, that were tested at a common set of conditions. The University of California Santa Barbara (UCSB) b-B database was analyzed using three-dimensional finite element calculation of the crack tip fields combined with a cleavage model calibrated to the local fracture properties of the Shoreham steel. This paper focuses on the possible significance of these results to the Master Curve standard as formulated in ASTM E 1921. The statistical scaling procedure to treat variations in B used in E 1921 was found to be reasonably consistent with the UCSB b-B database. However, constraint loss for three point bend specimens begins at a deformation level that is much lower than the censoring limit specified in E 1921. Unrecognized constraint loss leads to a nonconservative, negative bias in the evaluation of To, estimated to be typically on the order of −10°C for pre-cracked Charpy specimens.
