Numerical evaluation of multi-metric data fusion based structural health monitoring of long span bridge structures
Journal
Structure and Infrastructure Engineering
Date Issued
June 3, 2018
DOI
10.1080/15732479.2017.1350984
Abstract
This work focuses on structural health monitoring of long span bridges for damage detection. A feature
extraction level data fusion based damage isolation strategy is presented using multi-metric sensing. The
multi-metric sensing uses two types of sensors, namely strain sensors and accelerometers. The methodology
combines the advantages offered by each type of sensors, while at the same time overcomes their
limitations. The flexibility index method is applied and the flexibility matrices based on the strain and
displacement data are combined after performing co-ordinate transformation. A study has been carried
out on a simulated finite element model of the Great Belt East Bridge where realistic damage scenarios
like damage in the girder, breaking of hanger cables, pier settlement, and loss of cable pretension were
introduced on the structure. The study indicates that multi-metric sensing is indeed necessary as it reduces
the possibility of false detections and increases the sensitivity and robustness of the methodology.
extraction level data fusion based damage isolation strategy is presented using multi-metric sensing. The
multi-metric sensing uses two types of sensors, namely strain sensors and accelerometers. The methodology
combines the advantages offered by each type of sensors, while at the same time overcomes their
limitations. The flexibility index method is applied and the flexibility matrices based on the strain and
displacement data are combined after performing co-ordinate transformation. A study has been carried
out on a simulated finite element model of the Great Belt East Bridge where realistic damage scenarios
like damage in the girder, breaking of hanger cables, pier settlement, and loss of cable pretension were
introduced on the structure. The study indicates that multi-metric sensing is indeed necessary as it reduces
the possibility of false detections and increases the sensitivity and robustness of the methodology.