Scholarlink

<p>This research investigates the application of machine learning (ML) algorithms to predict the compressive strength of concrete in which copper mine tailings (CMT) are partially used to replace fine aggregate, as part of a strategy for managing copper mine waste. The aim is to investigate the predictive performance of ML models in estimating compressive strength and to evaluate how CMT impacts compressive strength. Ten ML algorithms are compared: adaptive boosting (AdaBoost), categorical boosting (CatBoost), decision tree (DT), extreme gradient boosting (XGBoost), gradient boosting regression tree (GBRT), light gradient boosting machine (LightGBM), random forest regression (RFR), and support vector regression (SVR) with three different kernels (linear, polynomial, and radial basis function (RBF)). Based on the parity plots, performance metrics, residual plots, and Taylor diagrams, CatBoost, XGBoost, GBRT, and LightGBM are suitable ML algorithms for predicting the compressive strength of CMT. The R<sup>2</sup> values of these algorithms for testing and training are 0.95 and 0.99, respectively. Based on the performance metrics ranking, CatBoost is the best model in this study. The CatBoost reliably predicts compressive strength even when replacing sand with CMT, which generally leads to lower strength. Of the nine input features evaluated, curing age was determined to be the most important feature influencing compressive strength, followed by the amount of cement. This research highlights the potential of CatBoost as an effective model for estimating the compressive strength of CMT with the potential impact to reduce experimental costs and support the adoption of CMT as a construction material for environmental sustainability.</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 Comparative Exploration of Machine Learning Techniques for Compressive Strength Prediction in Copper Mine Tailing Concretes

  • 期刊论文

  • 发表日期：

    2025年4月25日

查看全文
  • Eka Oktavia Kurniati,
  • Kudzai Musarandega,
  • Sefiu O. Adewuyi,
  • Angelina Anani,
  • Hee Jeong Kim

摘要

This research investigates the application of machine learning (ML) algorithms to predict the compressive strength of concrete in which copper mine tailings (CMT) are partially used to replace fine aggregate, as part of a strategy for managing copper mine waste. The aim is to investigate the predictive performance of ML models in estimating compressive strength and to evaluate how CMT impacts compressive strength. Ten ML algorithms are compared: adaptive boosting (AdaBoost), categorical boosting (CatBoost), decision tree (DT), extreme gradient boosting (XGBoost), gradient boosting regression tree (GBRT), light gradient boosting machine (LightGBM), random forest regression (RFR), and support vector regression (SVR) with three different kernels (linear, polynomial, and radial basis function (RBF)). Based on the parity plots, performance metrics, residual plots, and Taylor diagrams, CatBoost, XGBoost, GBRT, and LightGBM are suitable ML algorithms for predicting the compressive strength of CMT. The R2 values of these algorithms for testing and training are 0.95 and 0.99, respectively. Based on the performance metrics ranking, CatBoost is the best model in this study. The CatBoost reliably predicts compressive strength even when replacing sand with CMT, which generally leads to lower strength. Of the nine input features evaluated, curing age was determined to be the most important feature influencing compressive strength, followed by the amount of cement. This research highlights the potential of CatBoost as an effective model for estimating the compressive strength of CMT with the potential impact to reduce experimental costs and support the adoption of CMT as a construction material for environmental sustainability.