Theoretical Study of Ductile Fracture in Steel Structures in the Presence of Spatial Variability in Toughness

Abstract

Micromechanical or local models are increasingly used for predicting microvoid-growth–induced ductile fracture in structural steel components. Methods to calibrate and apply these models presume that both calibration specimens and prototypical components are spatially homogenous in terms of material toughness. This presumption conflicts with test data that show significant variability in material toughness of coupons extracted from a larger steel sample. Spatial variability of toughness in structural components has the potential to diminish deformation capacity due to the statistical size effect, which arises from weakest-link sampling. To examine this issue, two material representations are evaluated against a set of 32 experiments on two types of structural steel, using a maximum likelihood estimation–based approach. One approach represents the material as homogenous with random toughness (as implicitly assumed by prevailing methods), whereas the other represents the material as heterogeneous with spatially random toughness. No significant difference is observed between the two approaches in predicting test data. This suggests that the material is approximately homogenous over the sizes of the tested coupons, with spatial variability present only over larger length scales. To examine the potential effects of such variability, parametric finite-element studies are conducted on a prototypical fracture-critical detail. The results confirm the statistical size effect, such that larger details have lower deformation capacity. Current methods are not able to capture this effect when transferring fracture toughness from lab-scale to archetype-scale components. A parameter is proposed to facilitate this type of scaling.