Vibration-based damage localization and quantification in a pretensioned concrete girder using stochastic subspace identification and particle swarm model updating

Abstract

A popular method to conduct structural health monitoring is the spatio-temporal study of vibration signatures, where vibration properties are extracted from collected vibration responses. In this article, a novel methodology for extracting and analyzing distributed acceleration data for condition assessment of bridge girders is proposed. Three different techniques are fused, enabling robust damage detection, localization, and quantification. First, stochastic subspace identification is used as an output-only method to extract modal properties of the monitored structure. Second, a reduced-order stiffness matrix is reconstructed from the stochastic subspace identification data using the system equivalent reduction expansion process. Third, a particle swarm optimization algorithm is used to update a finite element model of the bridge girder to match the extracted reduced-order stiffness matrix and modal properties. The proposed approach is first verified through numerically simulated data of the girder and then validated using experimental data obtained from a full-scale pretensioned concrete beam that experienced two distinct states of damage. Results show that the method is capable of localizing and quantifying damages along the girder with good accuracy, and that results can be used to create a high-fidelity finite element model of the girder that could be leveraged for condition prognosis and forecasting.