A Critical Evaluation of Some Constitutive Models for Finite-Element Simulation of Structural Steel Components

Abstract

Currently, finite-element (FE) simulation is ubiquitous in engineering research and practice as a tool for solving a wide variety of problems. In steel structures, high-fidelity continuum FE (CFE) analysis is often used to characterize the mechanical behavior of structural members and geometrically complex components such as connections. Accurate description of the material constitutive behavior lies at the heart of these CFE simulations, and over the years, many phenomenologically-based rate-independent constitutive models have been proposed, and a select few have been incorporated in commercially available FE software. In this study, we examine the effect of constitutive model selection on the accuracy of FE analysis predictions of structural steel component responses. Models were selected to have a gradual progression in complexity; some are well-established, widely available in commercial FE software, and extensively used by researchers and practitioners, whereas others, despite being more advanced, have so far been limited in application. Given the difficulties in accurately accounting for extraneous variables common to full-scale experimental testing of realistic structural steel components such as, residual stresses, geometric imperfections, complex boundary conditions, and material inhomogeneities, a set of small-scale configured component experiments were designed to economically minimize these sources of uncertainty and establish a benchmark for critically examining the accuracy of model predictions. Data from these experiments, which included the accurate measurement of local strains that were as high as 75%, facilitated detailed comparisons between model predictions and experiment results, revealing both strengths and limitations of various models.