Uncertainty Quantification of State Estimation in Nonlinear Structural Systems with Application to Seismic Response in Buildings

Abstract

This paper presents the application of Bayesian filtering for the estimation and uncertainty quantification of interstory drifts and shears in partially instrumented buildings. We investigate the performance of four Bayesian filters: the extended, unscented, and ensemble Kalman filters, and the particle filter. The filters were compared in a simulation environment under ideal modeling and model error conditions. Computational efficiency, estimation accuracy and statistical error behavior of the filters was investigated. It was found that under ideal modeling conditions all the filters perform adequately. However, the filters exhibit significant sensitivity to parametric and nonparametric model errors. In the application studied, the sensitivity to parametric errors was nonsymmetric, with underestimation of model stiffness and yielding strength having a larger detrimental effect on the accuracy than overestimation of these parameters.