Numerical investigation of spalling in concrete exposed to fire

Abstract

Spalling of concrete subjected to elevated temperatures remains a controversial research area despite the pursuance of experimental investigation for more than a century. To date, experimental investigations have mostly been aimed at identifying spalling types, factors affecting spalling and methods to reduce spalling risk. However, numerous questions remain due to varied and sometimes contradictory results. To address these inconsistencies, this work aimed to investigate how the test methodologies themselves may influence the results and spalling behaviour. To do this, a numerical model was developed in ABAQUS and employed to simulate a consistent set of experiments conducted with different test configurations. The study was extended to investigate factors such as sample size, restraints (including mechanical, self-restraint via the so called ‘cold ring’ effect and applied load) and load intensity. Fundamentally, it was found that the boundary conditions of the test can have a significant influence on spalling behaviour due to their effect on the stress state in the samples. While mechanically restraining a sample can improve resistance to spalling, samples with the entire thickness restrained were found to have less resistance than those with only part of the thickness restrained. However, samples restrained through loading applied prior to heating tended to spall earlier than samples restrained mechanically. Increasing load intensity was found to worsen conditions for spalling while the position and eccentricity of loading can significantly vary the compressive regime in a sample. The influence of an unheated ‘cold ring’ around the sample was found to vary considerably with test configuration and in some cases had no effect on spalling. Sample thickness was found to be more significant, with thicker samples having less resistance to spalling than more flexible thinner samples. In conclusion, reflecting real conditions relating to size, loading and restraints in the test set up is strongly recommended to ensure representative results.