High-temperature Material Constitutive Models for Structural-Fire Analysis

Abstract

The applicability of three steel constitutive models was evaluated using finiteelement analyses and various member capacity equations. Three different hightemperature stress-strain models were compared: the model recently developed by the National Institute of Standards and Technology (NIST) [1], the Eurocode 3 model [2] and the model developed by Lie [3]. The testbed used in the analyses included twenty steel column tests and two restrained steel beam tests reported in the technical literature. The selected column tests reported buckling temperatures ranging from 500 oC to 700 oC and applied axial load ranging from 20 % to 65 % of the axial-load capacity at ambient temperature. Each reported test was analyzed in two different ways: (1) finite-element model was developed to predict the buckling temperature of the steel columns and response of the restrained steel beams in fire condition. (2) member capacity equations prescribed in Eurocode 3 and ANSI/AISC-360-10 [4] were used to compute the buckling temperature of the steel columns. Overall, the results indicate that all investigated material models give acceptable prediction of the buckling temperature of the steel columns and the behavior of restrained beams. The finite-element model with the NIST and the Lie material models predict the buckling temperature more accurately than that with the EC 3 material model. When the Eurocode column capacity equations were used, the buckling temperatures calculated using the NIST and the EC 3 models are more comparable with test results than those using the Lie model. It was also found that the current ANSI/AISC 360-10 Appendix 4 equation conservatively estimate the buckling temperature of the tested column specimens with difference of 20% on average. When the standard column equation in the Chapter E of ANSI/AISC 360-10 was used, both the EC 3 and the NIST models accurately predict the buckling temperature of the tested column specimen with difference less than 5% on average.