Experimental Study on Monitoring and Prediction of Earth-rockfill Dam Leakage Based on Infrared Thermography

The article delves into the critical issue of leakage in earth-rockfill dams, highlighting the limitations of traditional monitoring methods and the potential of Infrared Thermography (IRT) as a non-contact, indirect monitoring technique. It describes the design and construction of a leakage simulation platform to mimic real-world conditions and investigates the thermal imaging characteristics of dam leakage under various scenarios, including different dam slopes, leakage scales, and weather conditions. The study reveals distinct thermal imaging patterns for uncovered and vegetated slopes, with leakage appearing as low temperature anomalies during the day and high temperature anomalies at night. The article also explores the impact of rainfall on IRT monitoring, finding that daytime rainfall reduces the temperature difference between the reservoir water and the dam slope, making leakage detection more challenging, while nighttime rainfall increases this difference, enhancing monitoring effectiveness. Furthermore, the article introduces an innovative leakage prediction method based on the E3D-LSTM deep learning model, which captures the spatiotemporal characteristics of temperature fields from IRT time-series data. The model demonstrates promising prediction performance for both vegetated and uncovered slopes, offering valuable insights for anti-seepage design and emergency response strategies. The study concludes by discussing the practical implications of the findings and suggesting future research directions, such as integrating the proposed method with UAV technology and digital twin systems for comprehensive dam safety evaluation and early warning.