A computationally efficient Bayesian framework for structural health monitoring using physics-based models

Abstract

A Bayesian inference framework for structural damage identification is presented. Sophisticated structural identification methods, combining vibration information from the sensor network with the theoretical information built into a high-fidelity finite element model for simulating structural behaviour, are incorporated into the system in order to monitor structural condition, track structural changes and identify the location, type and extent of the damage. The methodology for damage detection combines the information contained in a set of measurement modal data with the information provided by a family of competitive, parameterized, finite element model classes simulating plausible damage scenarios in the structure. The computational challenges encountered in Bayesian tools for structural damage identification are addressed. Simulated modal data from the Metsovo Bridge are used to validate the effectiveness of the methodology. © Civil-Comp Press, 2015.