Non-contact measurements for the safety assessment of long-standing concrete dams: challenges and perspectives

Abstract

The safety of long-standing concrete dams still in operation is based on the analysis of data provided by accurate monitoring systems. Periodic measurements are performed by sensors positioned at specific sensitive locations such as the crest of the dam, or inside its body. Models of the structural behaviour support the interpretation of the information thus collected. Data-driven approaches based on machine-learning algorithms are often used. Once properly trained and tested, these techniques are effective in detecting anomalies in the structural response, possibly due to ongoing damage processes. The reliability of the outcomes provided by these potential warning systems can be significantly increased by integrating data-driven methods with physics-based simulation models. The realism and accuracy of the latter depend on the availability of displacement and temperature distributions over large portions of the exposed surfaces of the dam under study. These measurements can be provided by non-contact infrared and vision-based techniques, which are currently widely available. However, variations due to degradation processes may be low, and can be confused with experimental errors. Future advances in sensor technology are expected to improve measurement accuracy, while the integration of different sensing techniques could enhance the detectability and interpretation of even subtle changes.