A Hierarchical Filtering Approach for Online Damage Detection Using Parametric Reduced-Order Models

This chapter presents a hierarchical Bayesian framework for the system parameter identification of vibrating systems using spatially incomplete and noisy output-only response measurements. The parameters to be identified are treated as random variables, whose distributions are approximated by a finite number of evolving particles. For each realization of the parameters, an output-only Bayesian filter is employed for the unknown input and state estimation, creating thus a bank of filters that are recursively weighted, upon assimilation of the measurement information, and subsequently updated in order to narrow down the range of system parameters and converge to the target values.