Structural Health Monitoring with Artificial Neural Network and Subspace-Based Damage Indicators

In recent years, different structural health monitoring (SHM) systems have been proposed to assess the actual conditions of existing bridges and effectively manage maintenance programmes. Nowadays, artificial intelligence (AI) tools represent the frontier of research providing innovative non-invasive and non-destructive evaluations directly based on output-only vibration measures. This is one of the key aspects of smart structures of the future. In the current study, an artificial neural network (ANN) method has been proposed in order to perform damage detection based on subspace-based damage indicators (DIs) and other statistical indicators. A numerical case study example has been analysed with simulated damaged conditions. Based on a comparison between a reference situation and a new one, the greatest advantage in adopting these particular DIs is because they are able to point out significant changes, i.e. possible damage, without requiring a beforehand modal identification procedure, which may introduce further noise and modelling errors inside the traditional damage detection process.