STRUCTURAL DAMAGE DETECTION IN BRIDGES UNDER OPERATIONAL AND ENVIRONMENTAL VARIABILITY

Ensuring the safety and reliability of bridges, essential elements of civil infrastructure, requires precise assessment methods. Traditional structural health monitoring often associates changes in dynamic response with possible damage. However, for bridges, changes can also derive from operational factors like traffic loads or environmental influences, such as temperature and humidity. These variations, unrelated to structural integrity, complicate damage detection, as they can cause false alarms. To address this, a methodology designed to detect and localize bridge damage while accounting for these external factors is proposed. This approach relies on acceleration data from the Yonghe Bridge, a cable-stayed bridge in China, collected as part of a continuous monitoring effort. The non-stationary nature of these signals limits the effectiveness of the Fast Fourier Transform, prompting the use of Variational Mode Decomposition to separate the data into meaningful Intrinsic Mode Functions. Subsequently, instantaneous frequencies are derived through the Hilbert Huang Transform, identifying damage-sensitive features within the signal. Environmental and operational influences on these features are attenuated via Principal Component Analysis, a dimensionality reduction technique based on variance that enhances interpretability without significant data loss. For the final stage, statistical analysis selects critical features for a clustering process, applying the K-means Machine Learning algorithm to identify damage location. This comprehensive approach has shown a high degree of accuracy in identifying damage under varying traffic and environmental conditions, suggesting its applicability for structural health monitoring systems.