Scholarlink

<p>Ensuring the structural integrity and safety of foundations depends on the accurate prediction of elastic settlement. In this study, machine learning (ML) techniques are used to predict the elastic or immediate settlement of a rectangular footing in cohesionless soil. Elastic settlement occurs when an imposed load compresses the soil initially beneath a footing, causing a drop in volume and subsequent settling of the soil as the particles reorganize and compress under pressure. The aim of this study is to propose a high-performance machine learning model to predict settlement. This study uses a dataset of 200 foundation settlements to compare the <i>K</i>-nearest neighbor (KNN), multi-layer perceptron (MLP), and support vector regression (SVR) techniques. Five input parameters are considered, namely foundation width (<i>B</i>), foundation length (<i>L</i>), foundation depth (<i>D</i>), load intensity (<i>q</i>), and average SPT blow count (<i>N</i>), which are used to predict the output. The predictive power of the models is assessed using various performance parameters, such as the coefficient of determination (<i>R</i><sup>2</sup>), Willmott’s index (WI), A-20 index, variance account factor (VAF), scatter index (SI), root mean square error (RMSE), mean absolute error (MAE), and median absolute deviation (MAD). Additionally, other analyses are performed, including rank analysis, radar diagram, regression plot, reliability index, William’s plot, and error matrix, to assess the best predicting model. From this study, it is observed that the performance of SVR model is better due to its higher value of <i>R</i><sup>2</sup> (tr = 0.879, ts = 0.784, and overall = 0.861) and the least value of RMSE (tr = 0.79, ts = 0.068, and overall = 0.101) while predicting the elastic settlement of the footing. To examine the influence of different input parameters on the output, sensitivity analysis is performed, revealing that <i>q</i> is the most influential parameter among all the inputs followed by <i>B</i>, <i>L</i>, <i>D</i>, and <i>N</i>-value.</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 elastic settlement of rectangular footing using machine learning techniques

  • 期刊论文

  • 发表日期：

    2025年1月31日

查看全文
  • Rashid Mustafa,
  • Ankit Anshuman

摘要

Ensuring the structural integrity and safety of foundations depends on the accurate prediction of elastic settlement. In this study, machine learning (ML) techniques are used to predict the elastic or immediate settlement of a rectangular footing in cohesionless soil. Elastic settlement occurs when an imposed load compresses the soil initially beneath a footing, causing a drop in volume and subsequent settling of the soil as the particles reorganize and compress under pressure. The aim of this study is to propose a high-performance machine learning model to predict settlement. This study uses a dataset of 200 foundation settlements to compare the K-nearest neighbor (KNN), multi-layer perceptron (MLP), and support vector regression (SVR) techniques. Five input parameters are considered, namely foundation width (B), foundation length (L), foundation depth (D), load intensity (q), and average SPT blow count (N), which are used to predict the output. The predictive power of the models is assessed using various performance parameters, such as the coefficient of determination (R2), Willmott’s index (WI), A-20 index, variance account factor (VAF), scatter index (SI), root mean square error (RMSE), mean absolute error (MAE), and median absolute deviation (MAD). Additionally, other analyses are performed, including rank analysis, radar diagram, regression plot, reliability index, William’s plot, and error matrix, to assess the best predicting model. From this study, it is observed that the performance of SVR model is better due to its higher value of R2 (tr = 0.879, ts = 0.784, and overall = 0.861) and the least value of RMSE (tr = 0.79, ts = 0.068, and overall = 0.101) while predicting the elastic settlement of the footing. To examine the influence of different input parameters on the output, sensitivity analysis is performed, revealing that q is the most influential parameter among all the inputs followed by B, L, D, and N-value.