Comparative analysis of cyclic models for steel beams

Phenomenological models are used to simulate post-elastic behaviour of steel beams in the framework of concentrated plasticity. Many of them can model degradation phenomena occurring during cyclic loading, but these are often governed by model parameters without clear physical meaning. It follows that in this case parameter calibration must be performed by curve fitting of experimental responses. In this work, several numerical models (simple hysteretic, modified Ibarra-Medina-Krawinkler, Bouc-Wen and Sivaselvan-Reinhorn), implemented in widely used software packages, are calibrated against the results provided by an experimental programme involving cyclic and monotonic tests on open and closed crosssection beams, by employing a multi-objective optimisation methodology recently developed by the authors. The extensive calibration analyses carried out show that the most accurate model among those investigated is the Sivalselvan-Reinhorn model, which is able to provide realistic simulations of both monotonic and cyclic responses. Extension of the calibration procedure is proposed, which considers an additional objective related to the envelope curve of the cyclic response, and it is shown that this improvement adds robustness to the results. Finally, a preliminary regression analysis of the results, aimed at correlating degradation model parameters to geometric and mechanical properties of the member, is described.