Vibration-based damage detection and damage-induced stress redistribution of a real steel truss bridge

This study investigated the practicability of a damage detection approach using vehicle-induced bridge vibrations to a real steel truss bridge subjected to artificial damage. Stress redistribution was observed under both static and dynamic vehicle loadings, implying that vehicle-induced bridge vibrations, even measured at members away from the damaged one, may carry information about damage. The damage were detected if a set of candidate observations collected from an unknown condition do not follow the same pattern of the observations from the known intact condition. The pattern dissimilarity was evaluated by the Mahalanobis-Taguchi system, by taking damage indicators calculated from a certain sensor group of bridge vibrations as variables. Among three candidate damage indicators, Nair's damage sensitive feature was more sensitive to the artificial damage than dominant frequencies and damping ratios were. Damage localization was also feasible if sensors were densely deployed near the damaged member and adjacent sensors were grouped as variables.