Steel fracture modeling at elevated temperature for structural-fire engineering analysis

Cai, Wenyu
Journal Title
Journal ISSN
Volume Title

One of the key elements of performance-based structural-fire safety design is the ability to accurately predict thermal and structural response to fire. For steel structures, significant advances have been made in using finite element models for predicting the response of members, connections and entire structural systems exposed to fire. However, predicting the initiation and propagation of fracture of structural steel at elevated temperatures is still very difficult and uncertain using even the most advanced finite elements software. Fracture plays a critical role in the response of steel structures to fire, and is particularly important in connection response, where fracture often controls both strength and deformation capacity. While advances have been made in computational prediction of the initiation and propagation of fracture in steel at room temperature, much less is known at elevated temperature.
The objective of the research described in this dissertation was to evaluate the ability of existing ductile fracture models for metals to predict initiation and propagation of fracture in structural steel at elevated temperatures. The general finite element program Abaqus was used in this research to explore and evaluate various approaches for simulation of fracture. In the first part of this study, true stress-strain curves were developed for structural steel at ambient and elevated temperatures that extend to very large, post-necking strains. Then two different fracture criteria were studied for modeling steel fracture at ambient and elevated temperatures in Abaqus. These two fracture criteria are referred to as the ductile fracture criterion and the shear fracture criterion. Both predict the equivalent plastic strain at fracture as a function of the state of stress, most notably the stress triaxiality, but have different formulations and model parameters. Model parameters for each fracture criterion were estimated by a calibration process that involved developing finite element models of various tests reported in the literature of structural steel materials, members, and connections at ambient and elevated temperatures. To evaluate the capabilities and limitations of each model, a number of comparisons were made between tests of steel components that failed by fracture, and simulations of those tests. These evaluations were conducted for tests conducted at temperatures ranging from ambient up to 1000C.
Results of this work showed that the calibrated ductile fracture model was able to reproduce the experimentally observed behavior of tension coupons at elevated temperatures, all the way up through complete fracture. However, this same calibrated ductile fracture model was significantly less accurate in predicting the experimentally observed elevated temperature behavior of bolted steel connections. The model significantly overestimated the measured deformation capacity of the connections. This implies that the model overestimated the equivalent plastic strain at fracture for the states of stress developed in the regions of the bolted connections that experienced fracture. The calibrated shear fracture model, on the other hand, was capable of predicting the observed behavior of a wide range of bolted connection tests with reasonable accuracy. At any given temperature, the same shear fracture model parameters were able to reasonably predict the fracture of a variety of steel grades as well as high strength bolts. This suggests that the fracture model parameters may not be highly sensitive to changes in steel strength. Based on information in the literature and observations from this research, neither the ductile fracture model nor the shear fracture model is applicable across a full range of stress triaxiality values. The ductile fracture model appears to be most appropriate for predicting fracture under high levels of stress triaxiality, whereas the shear fracture model appears most appropriate for states of stress characterized by lower levels of stress triaxiality. The attempts at fracture simulations in this dissertation are based on limited experimental data and should be considered preliminary in nature. Far more work is needed to further develop these capabilities. Nonetheless, the numerous comparisons between simulations and experiments provided in this dissertation offer the hope that fracture behavior of steel connections and members at elevated temperatures can ultimately be simulated with confidence.