Bayesian Model–Data Fusion for Mechanistic Postearthquake Damage Assessment of Building Structures

Abstract

This paper presents a probabilistic framework for estimating seismic-induced damage in partially instrumented buildings. The proposed framework uses acceleration measurements at a limited number of stories and Bayesian filtering to estimate the response at all stories. The paper compares four Bayesian filters: the extended, unscented, and ensemble Kalman filters, and the particle filter. The estimated response throughout the building serves as input to a damage model that yields an estimate of structural damage and its uncertainty at all stories. The methodology is numerically verified in an elastoplastic 5-story shear building and in a 10-story inelastic moment-resisting frame under various types of model errors and minimal instrumentation. It was found that under ideal and mild modeling error conditions, the proposed methodology provides consistent estimates of damage and its uncertainty.