Uncertainty of bridge vibration properties and its consequence for damage identification

This study deals with the topic of bridge health monitoring based on identification of bridge vibration properties, which are influenced by uncertainties. Focus is laid on quantification of the uncertainties from continuous monitoring data. The determined uncertainties are then used in probabilistic evaluation of structural state. The purpose is to acquire besides the "best-match" structural state also the reliability of identification. A method for automated extraction of modal parameters and evaluation of their uncertainties is presented. It comprises clustering and statistical evaluation of Stochastic Subspace Identification system poles. The statistical properties of modal parameters are used to build a set of probabilistic variables sampled using Monte-Carlo Simulation. The generated sets are used for probabilistic damage detection, which uses a forward approach of finding the best match within a database of precalculated structural states. The presented study was performed using monitoring data acquired on a prestressed concrete box-girder bridge in course of 9 months of continuous data recording.