Development of AI-Driven Structural Health Monitoring Systems for Bridges

Thesis Abstract

The deterioration and aging of bridge infrastructure pose significant risks to public safety, economic stability, and transportation efficiency, necessitating the development of advanced, real-time structural health monitoring (SHM) systems capable of early damage detection and maintenance planning. Traditional SHM techniques often rely on manual inspections or sensor data analysis that lack the sophistication to interpret complex structural behaviors and provide prompt decision-making support, thereby underscoring the need for innovative, automated solutions driven by artificial intelligence (AI). This research aims to develop an AI-driven SHM system tailored for bridges that enhances the accuracy, efficiency, and predictive capabilities of existing monitoring frameworks. The specific objectives include designing a comprehensive data acquisition framework integrating diverse sensor types; developing machine learning algorithms for anomaly detection, damage localization, and prognosis; validating the system through field data collected from a selected sample of 20 operational bridges within a metropolitan region; and assessing the system’s performance relative to conventional SHM approaches. The research adopts a mixed-methods approach, combining qualitative evaluation of system design and quantitative analysis using empirical data. The study employs an experimental research design, focusing on deploying sensor networks on selected bridges equipped with accelerometers, strain gauges, and displacement sensors, connected to a centralized data acquisition platform. The population comprises bridge structures subjected to varying traffic loads and environmental conditions, with a stratified random sampling technique selecting 20 bridges based on age, traffic volume, and structural type to ensure representativeness. Data collection instruments include wireless sensor nodes, data loggers, and field inspection reports, with the collected data subjected to preprocessing, feature extraction, and normalization. The machine learning models—particularly convolutional neural networks (CNN) and support vector machines (SVM)—are trained and validated using a dataset of 10,000 labeled instances representing normal and damaged states. Model performance is evaluated through metrics such as accuracy, precision, recall, F1-score, and receiver operating characteristic (ROC) curves. The study employs regression analysis to examine relationships between sensor signals and structural health indices, while principal component analysis (PCA) reduces feature dimensionality. Expected findings indicate that the AI-driven SHM system achieves higher detection accuracy, faster response times, and enhanced damage localization capabilities compared to conventional threshold-based methods. The integration of machine learning algorithms is anticipated to improve predictive maintenance scheduling and extend bridge service life through early fault detection. The research contributes to knowledge by providing a validated, scalable AI framework adaptable to various bridge types and environmental conditions, and by advancing theoretical understanding of AI applications in civil structural health management—particularly through the application of the Theory of Technological Innovation Diffusion and the Structural Reliability Theory. This study concludes that deploying AI-powered SHM systems significantly enhances bridge safety management and operational efficiency. It recommends the widespread adoption of such systems in routine infrastructure maintenance, continuous real-time monitoring, and emergency response planning. Furthermore, it advocates for future research into integrating Internet of Things (IoT) technologies with AI frameworks to create fully autonomous monitoring platforms capable of autonomous decision-making and maintenance scheduling. Overall, this research underscores the transformative potential of artificial intelligence in engineering infrastructure resilience, providing a robust, data-driven foundation for sustainable transportation networks.