Vibration Monitoring of Historical Towers: New Contributions from Data Science
摘要
Deep neural networks are used to study the ambient vibrations of the medieval towers of the San Frediano Cathedral and the Guinigi Palace in the historic centre of Lucca. The towers have been continuously monitored for many months via high-sensitivity seismic stations. The recorded data sets integrated with environmental parameters are employed to train a Temporal Fusion Transformer network and forecast the dynamic behaviour of the monitored structures. The results show that the adopted algorithm can learn the main features of the towers’ dynamic response, predict its evolution over time, and detect anomalies.