A Bayesian state-space approach for damage detection and classification

Abstract

The problem of automatic damage detection in civil structures is complex and requires a system that can interpret sensor data into meaningful information. We apply our recently developed switching Bayesian model for dependency analysis to the problems of damage detection, localization, and classification. The model relies on a state-space approach that accounts for noisy measurement processes and missing data. In addition, the model can infer statistical temporal dependency among measurement locations signifying the potential flow of information within the structure. A Gibbs sampling algorithm is used to simultaneously infer the latent states, the parameters of state dynamics, the dependence graph, as well as the changes in behavior. By employing a fully Bayesian approach, we are able to characterize uncertainty in these variables via their posterior distribution and answer questions probabilistically, such as “What is the probability that damage has occurred?” and “Given that damage has occurred, what is the most likely damage scenario?”. We use experimental test data from two laboratory structures: a simple cantilever beam and a more complex 3-story, 2-bay structure to demonstrate the methodology.