Deformation mode classification in extended endplate connections and implications on hysteretic response

Determining the mode by which steel connections deform under rotational demands is essential for assessing damage, quantifying the associated losses, tuning design, and characterizing the connection’s cyclic behavior. In this paper, a classification model is developed to predict the deformation mode in extended endplate connections (EEPCs) as a function of their layout, material, and geometric properties. The model covers six modes inclusive of those expected to occur in either fully rigid or partial strength EEPCs. Such modes, and particularly interactive ones, can be challenging to predict using traditional mechanical or analytical methods. The classification model utilizes the Random Forest algorithm and is trained using a large dataset of experimental and simulation data to achieve a high accuracy larger than 95 %. Additionally, recommendations are provided for characterizing hysteretic phenomenological models depending on the deformation mode. This includes an empirical formula for defining the cyclic pinching parameters in EEPCs undergoing endplate bending. This aims to support system-level seismic simulations employing the lumped plasticity approach. • Classification model for the deformation mode(s) in extended endplate connections. • Quantification of the probability of occurrence of each possible deformation modes. • Classification model can be used in designing connections for better ductility/repairability. • Classification model can be used to assess connection damage states for loss quantification. • Calibrated pinching models can assist in system-level numerical simulations.