Parameter Estimation of a Shake-Table Tested Bridge Column with Bond-Slip Effect Using Stochastic Inference

Stochastic filtering strategies, such as the unscented Kalman filter (UKF), are widely used for system or damage identification problems in various fields. However, in civil engineering, it is rarely used to learn about unknown modeling aspects using experimental data, despite of many studies using simulated data. This paper uses the UKF to estimate parameters of a nonlinear finite element (FE) model of a reinforced concrete (RC) bridge column where bond-slip effect played a significant role. The bridge column was tested on a shake table at the University of California, San Diego, and was subjected to a suite of seismic input motions of varying intensities. Although perfect bonding is a common engineering assumption, the experimental observations indicated that the bond-slip effect should be taken into account due to high contribution to fixed-end rotation to column drifts. In this regard, a computationally efficient nonlinear FE model is developed in OpenSees using fiber-based beam-column elements with fixed-end rotation considered. The FE model is updated by minimizing the discrepancy between FE-predicted and measured response, leading to optimum estimate of the unknown model parameters (e.g., those for core concrete and bond-slip). The UKF is employed for nonlinear FE model updating after careful selection of critical model parameters based on a sensitivity study. Data from different tests with seismic excitations of various intensity levels are also considered. In sum, this study successfully applies the UKF to estimate unknown modeling aspects using experimental data and affirms that the bond slip effect in the tested bridge pier column has a significant impact in its dynamic response.