Scholarlink

<p>This study aims to develop a non-destructive approach to experimentally evaluate and predict the compressive strength of eco-friendly mortars containing plastic waste (PA), recycled concrete aggregate (RCA), and basalt fiber (BF) under elevated-temperature conditions using ensemble learning (EL) algorithms. In this scope, specimens from 64 different mixtures were produced and exposed to 22, 400, 600, and 800&#xa0;°C, yielding an experimental dataset of 256 data points. The dataset was split into 80% training and 20% testing. Bagging-based Bagging Regressor (BR) and Random Forest (RF) and Boosting-based AdaBoost Regressor (ABR), Gradient Boosting Regressor (GBR), Extreme Gradient Boosting Regressor (XGR) and Light Gradient Boosting Regressor (LGR) models were trained. Hyperparameters were optimized using GridSearchCV with five-fold cross-validation. According to the results, the GBR model provided the highest accuracy on the test data with a coefficient of determination (R²) of 0.980 and a root mean square error (RMSE) of 1.826&#xa0;MPa. The LGR model followed very closely with an R² of 0.979 and an RMSE of 1.829&#xa0;MPa. The SHAP (SHapley Additive exPlanations) analysis indicated that temperature was the most influential variable, and compressive strength decreased markedly with increasing temperature. At 800&#xa0;°C, the reference strength decreased from 48.89&#xa0;MPa to 1.31&#xa0;MPa. The increase in PA ratio also had a decreasing effect on strength; the strength of the specimens containing 22.5% PA was only 0.66&#xa0;MPa at 800&#xa0;°C. A graphical user interface (GUI) was developed to enable instant strength predictions based on temperature and mixture proportions, providing a practical non-destructive prediction tool.</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

A study on the evaluation of the compressive strength of eco-friendly composite mortars exposed to elevated temperature by ensemble learning based method

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月23日

查看全文
  • Yılmaz Yılmaz,
  • Safa Nayır,
  • Memduh Nas,
  • Vahiddin Alperen Baki

摘要

This study aims to develop a non-destructive approach to experimentally evaluate and predict the compressive strength of eco-friendly mortars containing plastic waste (PA), recycled concrete aggregate (RCA), and basalt fiber (BF) under elevated-temperature conditions using ensemble learning (EL) algorithms. In this scope, specimens from 64 different mixtures were produced and exposed to 22, 400, 600, and 800 °C, yielding an experimental dataset of 256 data points. The dataset was split into 80% training and 20% testing. Bagging-based Bagging Regressor (BR) and Random Forest (RF) and Boosting-based AdaBoost Regressor (ABR), Gradient Boosting Regressor (GBR), Extreme Gradient Boosting Regressor (XGR) and Light Gradient Boosting Regressor (LGR) models were trained. Hyperparameters were optimized using GridSearchCV with five-fold cross-validation. According to the results, the GBR model provided the highest accuracy on the test data with a coefficient of determination (R²) of 0.980 and a root mean square error (RMSE) of 1.826 MPa. The LGR model followed very closely with an R² of 0.979 and an RMSE of 1.829 MPa. The SHAP (SHapley Additive exPlanations) analysis indicated that temperature was the most influential variable, and compressive strength decreased markedly with increasing temperature. At 800 °C, the reference strength decreased from 48.89 MPa to 1.31 MPa. The increase in PA ratio also had a decreasing effect on strength; the strength of the specimens containing 22.5% PA was only 0.66 MPa at 800 °C. A graphical user interface (GUI) was developed to enable instant strength predictions based on temperature and mixture proportions, providing a practical non-destructive prediction tool.