Data normalization for the continuous monitoring of a steel truss bridge: A case study from the Italian railway line

Structural health monitoring is recognized as a powerful tool to assist bridge management. Continuous long-term monitoring of bridge structures presents several challenges, including the need for effective system design, robust sensors deployment, efficient data management, and comprehensive data analysis and interpretation. In the field of operational modal analysis, automatic tracking of bridge frequencies over time has been shown to be significantly influenced by temperature fluctuations. This effect is also observed in low-frequency sampled signals. To address these issues, the authors present a double-step procedure to effectively mitigate the influence of temperature on the estimated modal parameters and raw signals from displacement, strain and rotation transducers. The procedure is based on multiple linear regression, taking the measured temperatures as inputs, followed by low-pass filtering operations applied to the residuals through moving averages, leading to the creation of minimum detectable anomaly curves. The latter allow to establish quantitative relationships between filtering window lengths and detectable damage thresholds at specified confidence levels. The case study involves a railway steel truss bridge, where more than a year of data was collected through a permanent monitoring system. The monitoring layout includes a variety of sensors deployed to measure the structural response, as well as environmental and operational variables. A 15-month dataset demonstrates how temperature compensation effectively reduces signal variability, which is crucial for enhancing early-stage anomalies detection.