Scholarlink

<p>Industrial production has escalated with advancing technology and growing populations, leading to increased environmental damage and global environmental concerns. This research explores using eco-friendly pozzolans, which are waste products from industry, to reduce the amount of cement needed for improving the base layer under the road using artificial intelligence methods. This research highlights the utilization of pozzolan industrial waste products at 10–30% replacement levels as a sustainable option to diminish the quantity of cement used in ground improvement projects. Six different ML algorithms (Support Vector Regression, Random Forest and XGBoost) and statistical models (Ridge, LASSO and Elastic Net regression models) are used to estimate the unconfined compressive strength (UCS) of cylindrical samples made from crushed stone soil with various mixtures. The investigated relationship of the experimental data with UCS is consistent with the conducted exploratory data analysis, predictive analysis and literature. This approach recycles waste materials and contributes to mitigating the amount of cement that causes CO₂ emissions during its production with robust ML models. Although the dataset has 83 rows and unbalanced characteristics, XGBoost shows the best predictive performance, with MSE = 2.3340, RMSE = 1.5277, NRMSE = 0.1368, MAE = 1.2856, MAPE = 0.1480, and MASE = 0.4484 for the test dataset. These results demonstrate the importance of integrating machine learning techniques in geotechnical engineering to optimize material use and enhance sustainability. By leveraging advanced predictive models, construction practices can be further refined, and the environmental footprint of infrastructure development can be reduced.</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

Leveraging eco-friendly pozzolans and machine learning for reduced cement use in soil improvement of road base layers

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月4日

查看全文
  • Esra Uray,
  • Zarina Oflaz

摘要

Industrial production has escalated with advancing technology and growing populations, leading to increased environmental damage and global environmental concerns. This research explores using eco-friendly pozzolans, which are waste products from industry, to reduce the amount of cement needed for improving the base layer under the road using artificial intelligence methods. This research highlights the utilization of pozzolan industrial waste products at 10–30% replacement levels as a sustainable option to diminish the quantity of cement used in ground improvement projects. Six different ML algorithms (Support Vector Regression, Random Forest and XGBoost) and statistical models (Ridge, LASSO and Elastic Net regression models) are used to estimate the unconfined compressive strength (UCS) of cylindrical samples made from crushed stone soil with various mixtures. The investigated relationship of the experimental data with UCS is consistent with the conducted exploratory data analysis, predictive analysis and literature. This approach recycles waste materials and contributes to mitigating the amount of cement that causes CO₂ emissions during its production with robust ML models. Although the dataset has 83 rows and unbalanced characteristics, XGBoost shows the best predictive performance, with MSE = 2.3340, RMSE = 1.5277, NRMSE = 0.1368, MAE = 1.2856, MAPE = 0.1480, and MASE = 0.4484 for the test dataset. These results demonstrate the importance of integrating machine learning techniques in geotechnical engineering to optimize material use and enhance sustainability. By leveraging advanced predictive models, construction practices can be further refined, and the environmental footprint of infrastructure development can be reduced.