Scholarlink

<p>The current research proposes a machine learning (ML)-based framework to predict the undrained bearing capacity of strip footings above voids and to investigate the influence of key governing parameters. A standalone Extreme Gradient Boosting (XGBoost) model is first developed as a baseline predictor. To further enhance its performance, XGBoost is hybridized with four metaheuristic optimization algorithms: Particle Swarm Optimization (PSO), Slime Mould Algorithm (SMA), Gannet Optimization Algorithm (GOA), and Beluga Whale Optimizer (BWO) for hyperparameter tuning, resulting in four hybrid models: XGBoost-PSO, XGBoost-SMA, XGBoost-GOA, and XGBoost-BWO. The dataset comprises 404 randomly generated samples obtained from finite element simulations conducted using Plaxis 2D (V20). Each sample includes seven input variables: void width (W/B), vertical distance (Y/B), horizontal distance (X/B), void height (H/B), load eccentricity (e/B), inclination angle (i), and undrained shear strength of soil (Cu), while the undrained bearing capacity (qu) is considered as the output variable. The dataset was divided into 80% for training and 20% for testing. Model performance was evaluated using eight statistical indicators: Mean Absolute Error (MAE), Scatter Index (SI), Relative Absolute Error (RAE), Root Relative Squared Error (RRSE), Standard Deviation Error (SDE), Theil Inequality Coefficient (TIC), Index of Agreement (IA), and the Coefficient of Determination (R<sup>2</sup>). Comparative analysis demonstrates that the XGBoost-SMA and XGBoost-PSO hybrid models achieve superior predictive accuracy, whereas the XGBoost-GOA and standalone XGBoost models show the poorest performance.</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 undrained bearing capacity of strip footings above voids using hybrid machine learning models

  • 期刊论文

  • 发表日期：

    2026年3月16日

查看全文
  • Saloua Hamza,
  • Brahim Lafifi,
  • Asma Hamlaoui,
  • Ammar Rouaiguia

摘要

The current research proposes a machine learning (ML)-based framework to predict the undrained bearing capacity of strip footings above voids and to investigate the influence of key governing parameters. A standalone Extreme Gradient Boosting (XGBoost) model is first developed as a baseline predictor. To further enhance its performance, XGBoost is hybridized with four metaheuristic optimization algorithms: Particle Swarm Optimization (PSO), Slime Mould Algorithm (SMA), Gannet Optimization Algorithm (GOA), and Beluga Whale Optimizer (BWO) for hyperparameter tuning, resulting in four hybrid models: XGBoost-PSO, XGBoost-SMA, XGBoost-GOA, and XGBoost-BWO. The dataset comprises 404 randomly generated samples obtained from finite element simulations conducted using Plaxis 2D (V20). Each sample includes seven input variables: void width (W/B), vertical distance (Y/B), horizontal distance (X/B), void height (H/B), load eccentricity (e/B), inclination angle (i), and undrained shear strength of soil (Cu), while the undrained bearing capacity (qu) is considered as the output variable. The dataset was divided into 80% for training and 20% for testing. Model performance was evaluated using eight statistical indicators: Mean Absolute Error (MAE), Scatter Index (SI), Relative Absolute Error (RAE), Root Relative Squared Error (RRSE), Standard Deviation Error (SDE), Theil Inequality Coefficient (TIC), Index of Agreement (IA), and the Coefficient of Determination (R2). Comparative analysis demonstrates that the XGBoost-SMA and XGBoost-PSO hybrid models achieve superior predictive accuracy, whereas the XGBoost-GOA and standalone XGBoost models show the poorest performance.