Evaluating maximum inter-story drift ratios of building structures using time-varying models and Bayesian filters

This paper presents a method for evaluating the maximum inter-story drift ratios (IDR) of building structures using a time-varying model and Bayesian filters. During earthquakes, parameters of the time-varying model walked randomly according to Gaussian distributions to adapt to time-varying stiffness and time-varying damping of real building structures, and then the time-varying model and Bayesian filters were combined in the state space to estimate the maximum IDRs. Three Bayesian filters were investigated in this study: the unscented Kalman filter (UKF), second-order extended Kalman filter (sEKF), and particle filter (PF). The performance of the proposed estimation method, including accuracy, robustness, computational efficiency and uncertainty, was studied by using a numerical four-story shear model and three-story steel frame shaking table test. In addition, the benefits of using the time-varying model and the Bayesian filters were revealed by comparing estimated results from the time-varying model-based and time-invariant model-based estimation methods with calculated and measured results.