Scholarlink

<p>The settlement of embankments constructed on soft soil remains a critical challenge due to the low shear strength and high compressibility of such soils. Cement-mixed columns are widely employed to enhance soil stability and reduce settlement by increasing load-bearing capacity and accelerating consolidation. The present study evaluated various machine learning algorithms, including linear and non-linear regression models and artificial neural networks, for their effectiveness in predicting the settlement of railway embankments reinforced with cement-mixed columns. The models were examined using statistical metrics (mean absolute error, root mean square error, and coefficient of determination (<i>r</i><sup>2</sup>-score)) to examine their accuracy and reliability. While multivariate polynomial regression and random forest regression yielded satisfactory outcomes, the artificial neural network outperformed all models, achieving a mean absolute error of 0.004&#xa0;m, a root mean square error of 0.005&#xa0;m, and an <i>r</i><sup>2</sup>-score of 0.986, coming out as the most effective approach for predicting embankment settlement.</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 Settlement of Embankment on Soft Soil Using Machine Learning

  • 期刊论文

  • 发表日期：

    2025年8月14日

查看全文
  • Shraddha Sharma,
  • Ajay Pratap Singh Rathor,
  • Jitendra Kumar Sharma,
  • Deepak Bhatia

摘要

The settlement of embankments constructed on soft soil remains a critical challenge due to the low shear strength and high compressibility of such soils. Cement-mixed columns are widely employed to enhance soil stability and reduce settlement by increasing load-bearing capacity and accelerating consolidation. The present study evaluated various machine learning algorithms, including linear and non-linear regression models and artificial neural networks, for their effectiveness in predicting the settlement of railway embankments reinforced with cement-mixed columns. The models were examined using statistical metrics (mean absolute error, root mean square error, and coefficient of determination (r2-score)) to examine their accuracy and reliability. While multivariate polynomial regression and random forest regression yielded satisfactory outcomes, the artificial neural network outperformed all models, achieving a mean absolute error of 0.004 m, a root mean square error of 0.005 m, and an r2-score of 0.986, coming out as the most effective approach for predicting embankment settlement.