Leveraging advanced numerical calibration to filter out temperature effects on vibration-based monitoring data: application to the Mogadouro clock tower
Please login to view abstract download link
Vibration-based data-driven damage identification methods have proven to be cost-effective strategies for the preventive conservation of complex existing buildings, such as heritage structures. However, the extracted modal properties are strongly influenced by variations in environmental parameters, such as temperature, humidity, wind speed and direction. These factors can mask emerging anomalies in the data induced by damage, leading to significant delays in detection and warning. Filtering out environmental parameters’ effects from vibration-based data of historic masonry structures is particularly challenging as the underlying phenomena still need to be fully understood. Moreover, the current structural conditions, often characterised by pre-existing damage, make it difficult to acquire reliable baseline samples over an extended training period. This study aims to develop a physics-informed machine learning approach for cost-effective damage detection and early warning of a monitored structure by combining the capability of predicting the modal properties variations (such as the natural frequencies) under changing temperatures through a highly reliable Finite Element Model (FEM) calibrated to the experimental response of the structure. The relationship between temperature and modal properties, as evaluated through the FEM, is then used to normalise the monitoring data. This process filters out the effects of the environmental variation while magnifying the effects of damage, which is investigated through machine learning algorithms for classification purpose. The approach significantly reduces the duration of the training period needed to establish a reliable data-driven anomaly detection classifier by incorporating modelled scenarios. The procedure is validated by analysing a real case study, specifically the Mogadouro clock tower in Portugal.
