Assessing the structural conditions of large concrete dams involves collecting and processing a large amount of data on the response (e.g., in terms of displacements) to seasonal variations in external actions (typically, changes in temperature and water level). Measurements are usually provided by the instruments traditionally installed on the facility, while emerging vision-based technologies can provide supplementary information as needed. Artificial intelligence methodologies (mainly, machine and deep learning) trained on monitoring data can improve the quality of the gathered information, predict the expected structural response, and detect anomalies. However, locating and identifying the type and severity of potential damage remains a challenging task. This paper discusses the related difficulties with specific reference to the peculiarities of concrete dams, where the sensitivity of measurable quantities to degradation scenarios can be critical.
Bolzon et al. (Thu,) studied this question.