Experimental investigation of damage detection on a scale model of the Yeongjong Bridge

Recently, many important long-span bridges have been built in Korea, most of which are equipped with modern monitoring systems. As a part of the development of the Yeongjong Bridge monitoring system project, an experimental investigation of local damage detection on a 1/15 scale model of the Yeongjong Bridge floor system is performed. As many as 48 channels of dynamic strain data were collected, corresponding to strain gauge placements on the in-situ structure. Model excitation was induced via a series of impact hammer strikes. Twelve different damage cases are examined. Signal anomaly index (SAI), representing the amount of change in frequency response is calculated with measured dynamic strains and analyzed using a series of neural network classifiers to detect the presence and type of damage. Although saw-cuts representing small cracks were not reliably detectable, most other damage types, including sawcut cracks longer than 100mm, were found to be detectable by the neural networks trained with experimental data.