AI contribution to the monitoring and safety assessment of dams: Review and perspectives

Structural health monitoring strategies are essential to support dam safety assessment, especially as aging processes and changes in environmental conditions increase uncertainty in the long-term performance of these strategic infrastructures. In recent years, Artificial Intelligence (AI), particularly machine learning, has been increasingly applied to support the interpretation of monitoring data, driven by advances in algorithms as well as improvements in data acquisition and management. Most existing reviews on AI applications in dam engineering either adopt broad perspectives covering multiple disciplines or focus on specific techniques. In contrast, this article provides a surveillance-oriented narrative review, organizing the literature according to the main components of dam monitoring and the types of data involved, rather than exclusively by algorithm families. Three complementary areas are considered: continuous monitoring based on point-wise time series, vision-based inspection using RGB images, and full-field non-destructive testing based on optical techniques, with particular attention to infrared thermography and digital image correlation and the role of their outputs in supporting numerical model updating. For each area, the review discusses the surveillance tasks addressed, the types of methods currently adopted, and their main advances and limitations. Finally, open challenges and future research directions are outlined, with a focus on data integration, physical consistency, and interpretability in AI-assisted dam surveillance.