Prediction of performance of exterior beam-column connections with headed bars subject to load reversal

Given that both ACI 318-08 provisions and 352R-02 recommendations have been developed based on quite limited experimental data, an extensive database was assembled by Kang et al. [12], which contains most of the available test data of reinforced concrete exterior beam-column connections with headed bars subject to load reversal. In this study, the database has been further expanded by adding the recent data focusing on the investigation of design parameters of clear bar spacing and head size, and re-evaluated using a variety of statistical and empirical techniques. An effort has been made to find a statistical model linking quantitative design parameters and qualitative connection response. In this study, binomial logistic regression methodology has been applied. The statistical methodology quantifies the effect of each design parameter in determining the performance of the connection. A reliable and robust goodness-of-fit test, the log-likelihood ratio test, was performed to evaluate the developed logistic regression model. Finally, the recent connection data were used to validate the predictive capability of the developed statistical model.