Bond-slip model uncertainty quantification and effect on nonlinear behavior simulations of reinforced concrete columns

Abstract

Listen

Translate article

• Developing a new model for bonding properties based on the dataset compiled from the literature. • Quantifying the uncertainties of bonding properties through Bayesian linear regression. • Significant effect of variations of bonding properties on the nonlinear behavior simulation of RC columns with unhooked steel bars. • Need to consider the pertinent uncertainties of bonding properties in probabilistic dynamic analysis of RC columns. In advanced finite element (FE) modeling of reinforced concrete (RC) structures, realistic bond-slip behavior, i.e., imperfect bonding, at the steel–concrete interface needs to be considered with caution. This paper compiles an experimental database on bond-slip behavior from the literature, evaluates a variety of existing models for predicting bonding properties, and develops a new model to better characterize the bond-slip behavior using the experimental database. Recognizing variations in bonding properties, the authors also quantify the prevailing uncertainties as indicated by experimental data in the new model by Bayesian linear regression. To study the effect of bond-slip model uncertainties on the nonlinear behavior simulations of RC columns under static and dynamic loadings, the recently developed geometrically nonlinear fiber-based frame element capable of capturing bond-slip is used in this study together with a commonly-used bonding stress-slip pattern. To this end, three RC columns are firstly modeled with bonding properties determined from eight existing models and the newly developed one, showing that the simulated behaviors using different models can be different to various degrees. Next, the relative importance of the uncertainty in bonding properties is evaluated with comparison to that of the uncertainties in concrete/steel properties. The results indicate uncertainties associated with bonding properties can significantly alter the nonlinear behavior predictions of RC columns, and thus not negligible, particularly for probabilistic dynamic analysis of RC columns under large intensity excitation levels.