Enhancing Early-Stage Damage Detection in Bridges Through Advanced Feature Extraction Strategies

The safety, longevity, and operational efficiency of bridges are crucial as they are vital components of transportation networks. Traditional inspection methods often fail to detect subtle changes in bridge performance, potentially leading to serious failures and increased maintenance costs. To overcome these limitations, advanced Structural Health Monitoring (SHM) solutions are required to detect damage at early stages, enabling timely interventions and reducing maintenance expenses. However, implementing such solutions is challenging, particularly due to the lack of labeled data, which limits the use of supervised learning techniques. To address this issue, this work proposes an unsupervised, data-driven approach to detect the presence of damage in civil infrastructural systems. The proposed methodology relies on the extraction and implementation of different Damage-Sensitive Features (DSFs). Advanced signal processing methods, including wavelet transforms, principal component analysis, and independent component analysis, are employed, both independently and in combination, to identify DSFs that are sensitive to structural damage while being robust against environmental and operational variations common in bridge environments. The study involves a comparative evaluation of various DSF extraction techniques, assessing their effectiveness under different bridge scenarios characterized by varying loads and environmental conditions. Furthermore, feature fusion is explored by integrating multiple signal processing outputs and machine learning techniques to develop a comprehensive damage detection framework for bridges. This approach aims to enhance the accuracy of early detection and provide reliable assessments of bridge health. The proposed methodology is validated through extensive testing on a bridge structure subjected to a range of simulated damage scenarios and loading conditions reflective of real-world environmental and operational challenges. Results demonstrate significant improvements in the reliability and accuracy of bridge SHM systems, affirming the method’s effectiveness as a robust tool for continuous monitoring and maintenance of bridge infrastructure.