Scholarlink

<p>One-part geopolymers are promising sustainable alternatives to Portland cement, offering simpler handling through solid activators that only require water addition. To predict and confirm the compressive strength of one-part geopolymer mortar, as a greener alternative to conventional cement-based materials, the present work employs both experimental analysis and machine learning (ML). The mix was prepared with fly ash and ground granulated blast furnace slag (GGBS) and activated using solid alkalis including anhydrous sodium metasilicate and anhydrous sodium carbonate. A set of 135 samples were produced using varied mix ratios and subjected to curing periods of 3, 7, and 28 days. The compressive strength was determined through experiments, and the obtained dataset was utilized to train five ML models: Random Forest (RF), Gradient Boosting (GB), Decision Tree (DT), Artificial Neural Network (ANN), and Support Vector Regression (SVR). Cross- validation results revealed that RF achieved the highest predictive accuracy (R² = 0.82, RMSE = 20.07), closely followed by GB and DT, while SVR showed the weakest performance. SHAP analysis further confirmed that the Na₂SiO₃ content, curing age, and GGBS dosage were the most influential features, while excessive Na₂CO₃ and water content have shown decrement in the strength.</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

Analytical and experimental investigation on strength characteristics of one part geopolymer using machine learning models

  • 期刊论文

  • 发表日期：

    2025年11月27日

查看全文
  • R. Harika,
  • S. Venkateswara Rao,
  • Kolli Ramujee,
  • B. Keerthan

摘要

One-part geopolymers are promising sustainable alternatives to Portland cement, offering simpler handling through solid activators that only require water addition. To predict and confirm the compressive strength of one-part geopolymer mortar, as a greener alternative to conventional cement-based materials, the present work employs both experimental analysis and machine learning (ML). The mix was prepared with fly ash and ground granulated blast furnace slag (GGBS) and activated using solid alkalis including anhydrous sodium metasilicate and anhydrous sodium carbonate. A set of 135 samples were produced using varied mix ratios and subjected to curing periods of 3, 7, and 28 days. The compressive strength was determined through experiments, and the obtained dataset was utilized to train five ML models: Random Forest (RF), Gradient Boosting (GB), Decision Tree (DT), Artificial Neural Network (ANN), and Support Vector Regression (SVR). Cross- validation results revealed that RF achieved the highest predictive accuracy (R² = 0.82, RMSE = 20.07), closely followed by GB and DT, while SVR showed the weakest performance. SHAP analysis further confirmed that the Na₂SiO₃ content, curing age, and GGBS dosage were the most influential features, while excessive Na₂CO₃ and water content have shown decrement in the strength.