Scholarlink

<p>The van Genuchten and Fredlund–Xing empirical models are commonly used to construct Soil-Water Characteristic Curves (SWCCs) for evaluating soil suction, with each model exhibiting distinct strengths. This study employed a hybrid modelling approach, using an Artificial Neural Network (ANN) trained on data generated from the two SWCC models to improve curve-fitting and matric-suction prediction in a silty clay soil. To do so, the two models were first individually calibrated using experimental data. Matric suction-water content data were generated using the fitted models in the regions of the SWCC where each model excels, respectively. The data were then used to train an ANN to capture and integrate the complementary characteristics of the two models. The trained ANN reproduced the characteristic S-shaped SWCC and achieved high predictive accuracy, indicating effective integration of the underlying models. The proposed hybrid framework provides a physics-informed data-driven approach, demonstrating the potential of combining multiple physically based SWCC formulations to enhance predictive capability.</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

Prediction of Matric Suction in a Silty-Clay Soil Using an ANN-Stacked Computational Model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月28日

查看全文
  • Henok Marie Shiferaw

摘要

The van Genuchten and Fredlund–Xing empirical models are commonly used to construct Soil-Water Characteristic Curves (SWCCs) for evaluating soil suction, with each model exhibiting distinct strengths. This study employed a hybrid modelling approach, using an Artificial Neural Network (ANN) trained on data generated from the two SWCC models to improve curve-fitting and matric-suction prediction in a silty clay soil. To do so, the two models were first individually calibrated using experimental data. Matric suction-water content data were generated using the fitted models in the regions of the SWCC where each model excels, respectively. The data were then used to train an ANN to capture and integrate the complementary characteristics of the two models. The trained ANN reproduced the characteristic S-shaped SWCC and achieved high predictive accuracy, indicating effective integration of the underlying models. The proposed hybrid framework provides a physics-informed data-driven approach, demonstrating the potential of combining multiple physically based SWCC formulations to enhance predictive capability.