Full Program »

A mechanistic approach to seismic damage estimation of instrumented building structures using bayesian filtering

In instrumented buildings the measured dynamic response during a strong motion can be used to detect the presence of damage. The most direct approach is to use the measured response to identify changes in model parameters and relate those changes to structural damage. In this work we propose a different approach, namely, response measurements are used to estimate the evolution during the ground motion of mechanistic damage functions. A damage function is a response-based functional that yields a quantitative measure of the integrity of structural members or system after being subjected to a series of load/stress reversals. Damage functions are typically scaled such that a value of zero represents no damage and a value of one represents severe damage. Intermediate values have been calibrated using laboratory experiments.

The inputs to damage functions are response-based quantities such as dissipated hysteretic energy, drift displacements and ductility demands. To compute these quantities the complete dynamic response throughout the structure needs to be estimated using limited response measurements. The measured locations are limited, since instrumentation in buildings typically consists of a few accelerometers (~12-20) distributed throughout the structure. To perform the response estimation we propose the use of Bayesian filtering. To implement Bayesian filtering a nonlinear model, a stochastic model for the ground motion, and noise contaminated response measurements are necessary. In this study, four different Bayesian filters are considered: the extended, unscented and ensemble Kalman filters, and the particle filter. The performance of the various filters is compared in the context of a full-scale seven story building experiment subjected to damaging ground motions using a shake table.

Author(s):

Kalil Erazo    
University of Vermont
United States

Eric Hernandez    
University of Vermont
United States