Prediction of Inelastic Mechanisms Leading to Seismic Failure of Interior Reinforced Concrete Beam–Column Connections

Abstract

Inelastic mechanisms leading to failure in interior reinforced concrete beam–column (RCBC) connections, designed on the concept of strong column–weak beam philosophy, primarily result from failure of the joint region and yielding of longitudinal reinforcement in beams. In this manuscript, two novel easy-to-use probabilistic methodologies have been developed that can determine with sufficient accuracy the occurrence of either of these inelastic mechanisms leading to failure, given the geometric, material and loading parameters of an experimental investigation. One model was developed by using the relevance vector machine method, a machine learning methodology that uses a Bayesian formulation and results in a sparse representation. Another model was binomial logistic regression, which can relate the qualitative event of inelastic mechanism resulting in failure initiation with several experimentally obtained independent parameters. It can also quantify the relative importance of each of these independent parameters. Both methods show good predictive efficiency and can be utilized by a designer, engineer, or researcher to obtain a preliminary probabilistic estimate of inelastic mechanisms that lead to failure of interior RCBC connections. This manuscript also presents comparative evaluations of utilizing these two models.