Scholarlink

<p>This study presents a data-driven framework to optimize the mechanical and workability performance of geopolymer concrete (GPC) by refining key mix parameters liquid-to-binder ratio, NaOH molarity, SS/SH ratio, superplasticizer dosage, and curing temperature. Produced from industrial by-products such as fly ash and activated with alkaline solutions, GPC offers a sustainable alternative to cement-based concrete but lacks standardized mix design guidelines due to complex interactions among variables. A comprehensive dataset of 550 samples, comprising 300 literature data points and 250 experimental mixes, was developed to train and validate predictive models. The dataset covers a wide range of parameters: L/B ratio (0.24–0.8), fly ash content (250–530&#xa0;kg/m<sup>3</sup>), coarse aggregates (654–1567&#xa0;kg/m<sup>3</sup>), fine aggregates (318–817&#xa0;kg/m<sup>3</sup>), NaOH molarity (8–16&#xa0;M), SS/SH ratio (1.5–3.5), and curing temperature (27–100&#xa0;°C). Experimental results identified optimal mix conditions: L/B ratio of 0.55, 1.5% superplasticizer, NaOH molarity of 10&#xa0;M, SS/SH ratio of 2.0, and curing at 100&#xa0;°C, achieving a maximum compressive strength of 43.65&#xa0;MPa. To address the nonlinear relationships among variables, a hybrid Support Vector Machine optimized with a Genetic Algorithm (SVM + GA) was developed following Mutual Information–based feature selection. The model achieved R<sup>2</sup> = 0.9371, MAE = 0.0629, and RMSE = 0.2508 at 80% learning, outperforming Random Forest (R<sup>2</sup> = 0.9275) and ANN (R<sup>2</sup> = 0.8786). Statistical analysis (p &lt; 0.05) confirmed its significant improvement.</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 and multi-parameter optimization of geopolymer mixes using SVM-GA hybrid approach

  • 期刊论文

  • 发表日期：

    2025年11月12日

查看全文
  • Diksha,
  • Nerusupalli Dinesh Kumar Reddy,
  • Nirendra Dev,
  • Pradeep Kumar Goyal

摘要

This study presents a data-driven framework to optimize the mechanical and workability performance of geopolymer concrete (GPC) by refining key mix parameters liquid-to-binder ratio, NaOH molarity, SS/SH ratio, superplasticizer dosage, and curing temperature. Produced from industrial by-products such as fly ash and activated with alkaline solutions, GPC offers a sustainable alternative to cement-based concrete but lacks standardized mix design guidelines due to complex interactions among variables. A comprehensive dataset of 550 samples, comprising 300 literature data points and 250 experimental mixes, was developed to train and validate predictive models. The dataset covers a wide range of parameters: L/B ratio (0.24–0.8), fly ash content (250–530 kg/m3), coarse aggregates (654–1567 kg/m3), fine aggregates (318–817 kg/m3), NaOH molarity (8–16 M), SS/SH ratio (1.5–3.5), and curing temperature (27–100 °C). Experimental results identified optimal mix conditions: L/B ratio of 0.55, 1.5% superplasticizer, NaOH molarity of 10 M, SS/SH ratio of 2.0, and curing at 100 °C, achieving a maximum compressive strength of 43.65 MPa. To address the nonlinear relationships among variables, a hybrid Support Vector Machine optimized with a Genetic Algorithm (SVM + GA) was developed following Mutual Information–based feature selection. The model achieved R2 = 0.9371, MAE = 0.0629, and RMSE = 0.2508 at 80% learning, outperforming Random Forest (R2 = 0.9275) and ANN (R2 = 0.8786). Statistical analysis (p < 0.05) confirmed its significant improvement.