Scholarlink

<p>Recycled coarse aggregate concrete exhibits highly variable performance because recycled coarse aggregates (RCA) differ in morphology, surface texture, and microcracking. This study develops an integrated machine-learning and life-cycle assessment (ML–LCA) framework to quantify how RCA shape simultaneously influences compressive strength and environmental impacts. A 220-mixture dataset combining literature values and experimentally scanned aggregates was used to train three complementary models, Linear Regression (LR), Random Forest (RF), and Artificial Neural Network (ANN). LR served as the most accurate and stable predictor (R<sup>2</sup> = 0.919), followed by RF and ANN, and was therefore used to drive the LCA calculations. Morphology-dependent features such as angularity, sphericity, elongation, and fractal roughness were interpreted using SHAP, PDP curves, and Sobol global sensitivity analysis, confirming that mix parameters dominate strength variation while morphology contributes 20–40% through nonlinear interactions. Predicted strengths for elongated (34.7&#xa0;MPa), flat (33.4&#xa0;MPa), angular (34.8&#xa0;MPa), sub-angular (34.9&#xa0;MPa), and polished aggregates (36.1&#xa0;MPa), respectively, indicate that polished and sub-angular particles exhibit the highest performance. ML-predicted strengths were translated into dynamic cement-content adjustments and morphology-dependent crushing energy in OpenLCA, enabling environmental indicators (GWP, CED, ADP, AP, EP) to respond directly to particle geometry. Results show that optimized morphology can reduce cement demand by 3–6% and lower GWP by 4–8%, demonstrating that aggregate shape is a meaningful lever for both mechanical efficiency and sustainability. The proposed ML–LCA workflow provides a transparent, data-driven tool for designing RAC mixtures that balance mechanical performance with environmental impact.</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-based prediction of compressive strength and life cycle assessment of recycled aggregate concrete for sustainable and innovative infrastructure

  • 期刊论文

  • 发表日期：

    2026年3月3日

查看全文
  • Sujitha Arumugam,
  • Panruti Thangaraj Ravichandran

摘要

Recycled coarse aggregate concrete exhibits highly variable performance because recycled coarse aggregates (RCA) differ in morphology, surface texture, and microcracking. This study develops an integrated machine-learning and life-cycle assessment (ML–LCA) framework to quantify how RCA shape simultaneously influences compressive strength and environmental impacts. A 220-mixture dataset combining literature values and experimentally scanned aggregates was used to train three complementary models, Linear Regression (LR), Random Forest (RF), and Artificial Neural Network (ANN). LR served as the most accurate and stable predictor (R2 = 0.919), followed by RF and ANN, and was therefore used to drive the LCA calculations. Morphology-dependent features such as angularity, sphericity, elongation, and fractal roughness were interpreted using SHAP, PDP curves, and Sobol global sensitivity analysis, confirming that mix parameters dominate strength variation while morphology contributes 20–40% through nonlinear interactions. Predicted strengths for elongated (34.7 MPa), flat (33.4 MPa), angular (34.8 MPa), sub-angular (34.9 MPa), and polished aggregates (36.1 MPa), respectively, indicate that polished and sub-angular particles exhibit the highest performance. ML-predicted strengths were translated into dynamic cement-content adjustments and morphology-dependent crushing energy in OpenLCA, enabling environmental indicators (GWP, CED, ADP, AP, EP) to respond directly to particle geometry. Results show that optimized morphology can reduce cement demand by 3–6% and lower GWP by 4–8%, demonstrating that aggregate shape is a meaningful lever for both mechanical efficiency and sustainability. The proposed ML–LCA workflow provides a transparent, data-driven tool for designing RAC mixtures that balance mechanical performance with environmental impact.