Hydraulic and Hydroclimatic impact on dam seepage of civil and structural mechanisms with application of deep learning models

Seepage is a critical problem in earthfill dams which threatens the dam's stability and safety owing to extreme shifts in climate change with the rise in water intake in dams. To cope with this challenge dam monitoring is essential for structural stability and rehabilitation enhancement in earth dams with the integration of a deep learning approach. This research presents a novel approach for evaluating the dam seepage by using a Recurrent Neural Network and its associated co-variant to predict seepage at multiple location in earth fill Tarbela dam. Short-term peak seasonal hydraulic and hydro climatological monitoring data was used from Pakistan's Earth and Rockfill Tarbela dam over the period from 2014 to 2020. The results demonstrate that, compared to other models, the proposed model efficiently predicts the extent of seepage in the dam. This study highlights the importance of considering historical correlations in seepage data analysis, providing significant insights to stakeholders regarding the most effective utilization of data resources for monitoring purposes, and provide the idea of digitization of the earth fill dam monitoring system in Pakistan with integrated DL algorithms. •Deep Learning has become more prevalent in accurately modeling the behavior of dams.•DL methodologies for monitoring dam operations and decision-making is often hindered by models.•Recurrent Neural Network and its associated co-variant techniques to detect instances of seepage.•The proposed model effectively predicts the extent of seepage occurring in the dam.•This study focuses on the model's ability to accurately identify trends for anomaly identification.