Methodology for the assessment of bolted connections based on the finite element method and data mining techniques

Résumé

Connections between beams and columns are one of the most critical elements in steelstructures due to their highly nonlinear behaviour. The failure of a joint would producethe collapse of the whole structure so further details are needed to study geometricaldiscontinuities, local yielding, contact or stress concentrations.Nowadays, bolted connections are assessed through the component method which isincluded in the Eurocode 3. This mechanical method provides step-by-step analyticalformulae but do not take into account some of the nonlinearities embodied in the boltedconnections. In this paper, an alternative methodology based on the Finite ElementMethod (FEM) and Data Mining (DM) techniques is presented to predict the fullynonlinear behaviour of bolted connections. This methodology takes advantage of theFEM’s capacity but it is capable to reduce the computation time in a simplifiedmathematical model by means of the DM techniques.Finally, three cases study focus on the lap joint, the T-stub and the extended end-platebeam-to-column connections have been carried out by EDMANS group in order toapply this novelty methodology. Predictive models have been developed and validatedagainst experimental tests and the results show a high correlation between FEsimulations and the predictive models.