Prediction of the shear strength of RC beam-column joints using new ANN formulations

Abstract

Beam-column joint is a key part for reinforced concrete (RC) frame building structures under the impact of severe lateral actions of seismic or wind loading. Prior works indicate that there is lacking in the direction of formulation of new mathematical models to predict RC joints strength by artificial neural network (ANN) considering influential design parameters. Present endeavor is devoted to establish new empirical ANN models to estimate shear strength and failure mode for RC exterior and interior beam-column joints. These models were relied on training, testing and validation of a database of over 200 experiments, for the performance of RC joints, using ANN. More than 20 parameters, related to the geometry and loading of the joints, were selected as inputs for these models; but four effective inputs were used in the formulation of the final ANN models. The outputs for ANN models were compared to those counterparts of the commonly utilized models given in codes and other recent studies. This calculation confirms the reliability of current ANN mathematical models and their outstanding performance. Moreover, comparison shows that the proposed models were precise in including the effective independent variables in these models and avoiding the shortcoming of previous models.