While sophisticated simulation techniques are regularly applied in natural hazards engineering, they are limited in ability to capture several important failure mechanisms. A significant limitation is the capability to reliably simulate fracture in structural steel weldments. Predicting fracture in weldments is inherently challenging due to various factors. Moreover, when weldments do not have a sharp, pre-existing flaw, or are subjected to large-scale yielding or earthquake-induced Ultra Low Cycle Fatigue (ULCF – characterized by few cycles of large strain) – all of which occur frequently in modern buildings, conventional fracture and fatigue mechanics are invalidated. Given that fracture often initiates in the weld region, this represents a major obstacle to effective structural performance assessment. This is addressed this through a coordinated program of (1) physical tests and fractography on thermomechanically generated samples of various weld microstructures for discovery of fracture micromechanisms, (2) formulation of new continuum-based fracture criteria for these different microstructures, (3) numerical implementation, and (4) calibration and validation of the developed framework using laboratory testing. A key research challenge is to effectively upscale local models, considering spatial variation of microstructure and interaction between mechanisms to predict fracture at the component/structure scale, while also characterizing the uncertainty in these predictions.