Scholarlink

<p>Dam break events pose serious threats to human life and infrastructure, making their accurate modeling a critical area of research in hydraulic and civil engineering. This study proposes a novel hybrid machine learning approach combining the CatBoost algorithm with the Hunger Games Search (HGS) optimization technique to predict free surface water levels resulting from dam break scenarios. A laboratory dataset collected from a 6-m flume equipped with 16 measurement points and 32 scenarios varying in downstream water level (Dwl) and upstream sediment depth (Sd) was used for training and validation. The model’s predictive performance was compared against traditional numerical simulation methods, including Eulerian approaches and the Volume of Fluid (VOF) technique implemented within the finite volume framework. Evaluation metrics such as RMSE and MAE consistently demonstrated that the CatBoost-HGS hybrid model outperforms conventional techniques, achieving an R² of 0.9952 and reducing prediction errors by More than 70% in several scenarios. The results confirm that the proposed approach delivers superior accuracy in modeling dam break hydrodynamics, especially under varying sediment conditions, offering a reliable, efficient, and Low-cost alternative to traditional computational methods.</p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sensitivity Assessment and Comparative Analysis of Machine Learning and Numerical Models for Predicting Dam Break-Induced Water Levels

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Hamed Khajavi,
  • Amir Rastgoo,
  • Fariborz Masoumi

摘要

Dam break events pose serious threats to human life and infrastructure, making their accurate modeling a critical area of research in hydraulic and civil engineering. This study proposes a novel hybrid machine learning approach combining the CatBoost algorithm with the Hunger Games Search (HGS) optimization technique to predict free surface water levels resulting from dam break scenarios. A laboratory dataset collected from a 6-m flume equipped with 16 measurement points and 32 scenarios varying in downstream water level (Dwl) and upstream sediment depth (Sd) was used for training and validation. The model’s predictive performance was compared against traditional numerical simulation methods, including Eulerian approaches and the Volume of Fluid (VOF) technique implemented within the finite volume framework. Evaluation metrics such as RMSE and MAE consistently demonstrated that the CatBoost-HGS hybrid model outperforms conventional techniques, achieving an R² of 0.9952 and reducing prediction errors by More than 70% in several scenarios. The results confirm that the proposed approach delivers superior accuracy in modeling dam break hydrodynamics, especially under varying sediment conditions, offering a reliable, efficient, and Low-cost alternative to traditional computational methods.

Graphical Abstract