Author(s): Ahmed Emara; Sameh Kantoush; Mohamed Saber; Tetsuya Sumi And Emad Mabrouk
Linked Author(s): SAMEH KANTOUSH, Tetsuya Sumi, Mohamed Saber
Keywords: Invert Abrasion Susceptibility Model Abrasion Prediction Abrasion Mapping Event-Based Data Controlling Factors
Abstract: A vital engineering intervention for sediment management in reservoirs involves the use of Sediment Bypass Tunnels (SBTs). By channeling sediment-laden flows around dams, these tunnels mitigate sediment deposition, extending reservoirs' operational lifespan and preserving downstream riverine ecosystems. However, the operation of these tunnels, which frequently convey high-velocity flows, results in significant abrasion damage, particularly on tunnel floors. Forecasting this abrasion damage is crucial for the maintenance and long-term viability of SBTs, especially during flood events when sediment transport is intensified. This study aims to evaluate the predictive capabilities of advanced machine learning models that map abrasion susceptibility across tunnel floors, focusing on the Koshibu SBT, a 4 km tunnel with straight and curved sections. Specifically, this research explores how successive flood events influence Abrasion Susceptibility Mapping (ASM) within these models. The findings indicate a strong correlation between discharge magnitude and model performance in straight sections, whereas curved sections remain relatively unaffected. Notably, the model accuracy and correlation coefficient (R) values remained stable at approximately 80% and 67% when individual flood events were excluded, underscoring the necessity of event-based data for accurate post-event damage assessments. These results highlight the value of incorporating event-based data into machine learning models to enhance damage prediction in flood-prone structures.
Year: 2025