Date of Award
1-1-2025
Document Type
Dissertation
Degree Name
Ph.D. in Engineering Science
First Advisor
Ahmed Al-Ostaz
Second Advisor
Yacoub Najjar
Third Advisor
Hakan Yasarer
School
University of Mississippi
Relational Format
dissertation/thesis
Abstract
Flooding is one of the most destructive natural disasters in the United States, often resulting in catastrophic damage to communities and infrastructure. The primary flood-protection structures, earthen dams and levees, are critical for mitigating these impacts. In Mississippi alone, there are over 2,800 dams and 3,750 miles of levees along the Mississippi River and its tributaries. Of this, nearly 2,200 miles are along the mainstem Mississippi River, with the remaining levees situated along backwaters, tributaries, and floodways. Understanding the erosion behavior of these structures during overtopping events is essential for accurate failure prediction and risk assessment. This study focuses on predicting two key hydraulic erosion parameters for levee soils: Critical Shear Stress (τc) and the Erodibility Coefficient (kd). These parameters are fundamental in simulating levee breach processes using geotechnical-hydraulic software such as DL-Breach. An artificial neural network (ANN) model was developed to estimate τc and kd using a limited set of basic geotechnical properties, including soil texture, natural water content, and dry density. The study considers two ANN-based models trained on 39 datasets and 146 datasets, respectively, with the primary difference being the input properties and the experimental methods used to determine τc and kd. The optimal ANN structure was selected based on the highest coefficient of determination (R²) and the lowest mean absolute relative error (MARE) and average squared error (ASE). The best-performing model achieved an R² of 97.5%, indicating a strong correlation between predicted and experimental values. The predicted values of τc and kd are then integrated into levee failure simulations to generate breach hydrographs, allowing users to assess potential flood impacts based on site-specific input parameters. To enhance accessibility and practical application, the ANN model is integrated into DSS-WISE, a decision-support system for flood risk assessment. This integration allows users to input basic geotechnical properties and levee geometry, without requiring direct expertise in ANN modeling. DSS-WISE then utilizes the trained ANN model to estimate breach hydrographs and breach width, providing a streamlined approach for rapid and reliable levee failure analysis. This research advances levee breach modeling by offering a user-friendly, data-driven tool for flood risk management and decision-making.
Recommended Citation
Saleh, Haitham O., "Data-Driven Modeling of Levee Failure: A State-of-the-Art Approach for Integrating Artificial Intelligence Techniques with Hydraulic Erosion and Geotechnical Characterization" (2025). Electronic Theses and Dissertations. 3377.
https://egrove.olemiss.edu/etd/3377