Identification of Structural Damage in a Vehicular Bridge using Artificial Neural Networks

This article presents the application of artificial neural networks (ANNs) for the structural damage detection to bending in the girders of a vehicular bridge. An analytical model of the bridge was developed to generate 12,800 damage scenarios, in which the flexural stiffness of the elements was modified to simulate the damage. Such rigidities were used as output data for the network, while the modal strain energy differences were used as input data. To verify the NNs generalization capability in presence of noise in the measurements, four levels of noise were analyzed (2.5%, 5.0%, 7.5%, and 10.0%). It was observed that the developed NNs model is able to predict with high accuracy the location and severity of the damage in the studied bridge.