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

This study introduces an intelligent monitoring and prediction method for earth-rockfill dam leakage based on Infrared Thermography (IRT). To this end, a leakage simulation platform for earth-rockfill dam is designed and constructed to explore the thermal imaging characteristics and evolution process of dam leakage, and the impacts of various factors are evaluated by setting different slope surfaces, leakage scales, and weather scenes. Additionally, a leakage prediction method for earth-rockfill dams driven by IRT and image prediction model is proposed, offering a new approach to address the spatial limitations of traditional point-based leakage prediction methods. The E3D-LSTM network is utilized to capture the spatiotemporal features of temperature fields from infrared time-series data of dam leakage, enabling the prediction of future leakage trends. The results indicate that IRT can intuitively and accurately monitor leakage in earth-rockfill dams across most scenes. Under both the dynamic and transfer prediction strategies, the prediction results of earth-rockfill dam leakage reflect the distribution and development trend of leakage in spatiotemporal dimension.