Scholarlink

<p>The manufacturing of construction material which is sustainable in nature plays an important role in reducing the environmental impact related to Portland cement production and also promoting the principles of circular economy. The present study investigates a waste-glass-based cementless mortar, where waste glass powder played the role of primary binder component to avoid conventional usage of cement. The performance of the sustainable material was efficiently and optimally predicted employing the interpretable machine learning (ML) models. The key performance indicators namely compressive strength, flexural strength, and drying shrinkage were predicted by developing Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) ML models utilizing the experimental data. The evaluation of the models was rigorously conducted through train–test splitting and five-fold cross-validation, with the use of metrics such as the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Even though both ML models predicted with high accuracy, XGBoost models was observed to be marginally superior in terms of generalization performance for all the output parameters. Among the developed models, XGBoost demonstrated superior predictive capability with R² values of 0.996, 0.998, and 0.999 for compressive strength, flexural strength, and drying shrinkage, respectively, along with significantly reduced RMSE values. The enhancement in transparency and physical interpretability was obtained employing SHapley Additive exPlanations (SHAP) for studying the effect of input variables individually. The results focus on the usefulness of interpretable ML models as dependable decision-support tools for the design and optimization of low-carbon, waste-derived cementless mortar systems. The suggested framework offers a developed sustaianable material, maintaining transparency and minimizes the efforts and time spent on experimental work.</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

Predictive modeling of waste glass cementless mortar performance through interpretable machine learning approaches

  • 期刊论文

  • 发表日期：

    2026年4月14日

查看全文
  • K. G. Shwetha,
  • B. B. Das

摘要

The manufacturing of construction material which is sustainable in nature plays an important role in reducing the environmental impact related to Portland cement production and also promoting the principles of circular economy. The present study investigates a waste-glass-based cementless mortar, where waste glass powder played the role of primary binder component to avoid conventional usage of cement. The performance of the sustainable material was efficiently and optimally predicted employing the interpretable machine learning (ML) models. The key performance indicators namely compressive strength, flexural strength, and drying shrinkage were predicted by developing Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) ML models utilizing the experimental data. The evaluation of the models was rigorously conducted through train–test splitting and five-fold cross-validation, with the use of metrics such as the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Even though both ML models predicted with high accuracy, XGBoost models was observed to be marginally superior in terms of generalization performance for all the output parameters. Among the developed models, XGBoost demonstrated superior predictive capability with R² values of 0.996, 0.998, and 0.999 for compressive strength, flexural strength, and drying shrinkage, respectively, along with significantly reduced RMSE values. The enhancement in transparency and physical interpretability was obtained employing SHapley Additive exPlanations (SHAP) for studying the effect of input variables individually. The results focus on the usefulness of interpretable ML models as dependable decision-support tools for the design and optimization of low-carbon, waste-derived cementless mortar systems. The suggested framework offers a developed sustaianable material, maintaining transparency and minimizes the efforts and time spent on experimental work.