Resilience in Data-Driven Monitoring Adapting to Changing Environmental Factors in Infrastructural System

Data-driven monitoring strategies are essential for designing, operating, and maintaining infrastructure. These strategies use historical data to model patterns and relationships between system variables and environmental factors, such as temperature and humidity. However, climate change is altering these factors, causing a divergence between past data and current conditions, which can undermine monitoring strategies and lead to inaccurate predictions and increased risks. This research proposes a data-driven monitoring framework to analyze and forecast structural responses to environmental changes. The approach integrates system identification methods, Gaussian Process Regression (GPR), and climate projections. Structural behavior is characterized in terms of modal parameters using automated time-domain system identification techniques. These methods identify dynamic properties such as frequencies, damping ratios, and mode shapes directly from vibration data collected during monitoring. Changes in these parameters may indicate environmental variations or potential structural damage. The study employs GPR, a flexible statistical model that captures complex, non-linear relationships while providing predictive uncertainty estimates. Sparse Gaussian Processes are used to reduce computational costs while preserving accuracy. The kernel function is designed to reflect data smoothness and temporal correlations. To assess long-term climate effects on structures, Representative Concentration Pathways (RCPs) are incorporated, modeling future greenhouse gas concentration scenarios from low-emission (RCP2.6) to high-emission (RCP8.5). By integrating these projections with GPR, the study enables long-term assessments of structural behavior under evolving climate conditions. The proposed methodology is applied to the Chillon Viaduct in Switzerland. Using acceleration data from sensors, the study analyzes the bridge’s natural vibration frequencies to understand how temperature and time affect its behavior. The dynamic responses in terms of frequencies obtained from real acceleration data collected from 2017 are correlated with the temperature information measured on the bridge during the same period. This correlation allows the understanding and observation of short-term daily fluctuations and long-term trends. GPR models learn this relationship and predict future frequency variations driven by increasing temperatures, capturing trends and uncertainties. Results show that higher temperatures consistently reduce natural frequencies, likely due to material softening. To estimate future impacts, the study leverages the CH2018 model, which provides temperature projections for Switzerland up to year 2100. By integrating these projections, the research assesses potential long-term climate effects on the viaduct’s vibrations. This study highlights SHM’s power in preparing infrastructure for a warmer future. By predicting climate impacts early, engineers can prioritize repairs and ensure long-term safety. This approach offers a new point of view for protecting critical structures worldwide as temperatures continue to rise.