Scholarlink

<p>This study explores fly ash-involved geopolymer concrete (GPC) using silica fume (SF) (0–60&#xa0;kg/m³) and steel slag (SS) (0–720&#xa0;kg/m³) as additional industrial byproducts to assess mechanical behavior. Sixteen mix proportions were systematically designed by varying fly ash (340–400&#xa0;kg/m³), SF, and SS as partial replacements for natural coarse aggregate, activated using a constant 12&#xa0;M NaOH and Na₂SiO₃ alkaline solution. Experimental evaluation of hardened specimens was done on bulk density, compressive strength (CS), split tensile strength (TS), flexural strength (FS), and ultrasonic pulse velocity (UPV) at 7 days and 28 days of ambient curing. A mix of 7 with 5% SF and 40% SS replacement gave the optimum mechanical performance with a 28-day CS of 44&#xa0;MPa, TS of 4.6&#xa0;MPa, FS of 7.0&#xa0;MPa, and UPV of 4.9&#xa0;km/s, which meets M40 grade requirements as per IS 456:2000. The Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) algorithms were used to estimate the output properties, with KNN proving to be more accurate at all the parameters (R² = 0.949, RMSE = 8.57&#xa0;MPa in the case of CS). The most significant predictors were observed to be SS and natural coarse aggregate (feature importance analysis). The proposed ML framework reduces experimental dependency by approximately 80%, offering an efficient and sustainable approach to GPC mix design optimization.</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

Machine learning-driven prediction of mechanical properties in fly ash-based geopolymer concrete incorporating silica fume and steel slag

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • S. R. Hemalatha,
  • M. Sreedevi,
  • Bhukya Govardhan Naik,
  • Nakkeeran Ganasen,
  • R. Nidhya

摘要

This study explores fly ash-involved geopolymer concrete (GPC) using silica fume (SF) (0–60 kg/m³) and steel slag (SS) (0–720 kg/m³) as additional industrial byproducts to assess mechanical behavior. Sixteen mix proportions were systematically designed by varying fly ash (340–400 kg/m³), SF, and SS as partial replacements for natural coarse aggregate, activated using a constant 12 M NaOH and Na₂SiO₃ alkaline solution. Experimental evaluation of hardened specimens was done on bulk density, compressive strength (CS), split tensile strength (TS), flexural strength (FS), and ultrasonic pulse velocity (UPV) at 7 days and 28 days of ambient curing. A mix of 7 with 5% SF and 40% SS replacement gave the optimum mechanical performance with a 28-day CS of 44 MPa, TS of 4.6 MPa, FS of 7.0 MPa, and UPV of 4.9 km/s, which meets M40 grade requirements as per IS 456:2000. The Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) algorithms were used to estimate the output properties, with KNN proving to be more accurate at all the parameters (R² = 0.949, RMSE = 8.57 MPa in the case of CS). The most significant predictors were observed to be SS and natural coarse aggregate (feature importance analysis). The proposed ML framework reduces experimental dependency by approximately 80%, offering an efficient and sustainable approach to GPC mix design optimization.