This study aims to optimize the compressive strength (CS) of eco-concrete mixes that contain fly ash (FA), recycled concrete aggregate (RCA), and supplementary cementitious materials (SCMs). Using 324 datasets, a multivariate predictive modeling framework was created to guarantee excellent quality and consistency in production processes. CS optimization is crucial as a performance measure for industrial manufacturing of these alternatives. This research aims to develop a novel combination of designs and testing methods to improve the compressive strength of sustainable building materials. By enhancing their construction sustainability, these materials can support the sector's sustainability and meet the structural needs of various applications. This study employs multivariate predictive modeling to assess important variables systematically. This approach enables a comprehensive analysis of the various factors that interact to predict outcomes. Researchers can better understand the complex relationships by considering multiple variables simultaneously. Analytical models were built using 324 datasets collected from earlier research and nine input factors, including curing age, RCA replacement ratio, and W/C ratio, to assess the performance of four regression models: power equation (PE), non-linear (NLR), linear (LR), and multilinear (MLR). The independent variables are cement, 40 to 635 kg/m3; SCMs, 0 to 592 kg/m3; W/CMs, 0.25 to 0.45; fine aggregate (FA), 676 to 1033 kg/m3; coarse aggregate (RCA), 478 to 890 kg/m3; superplasticizer (SP), 1.21 to 8.60 kg/m3; recycled content, 0 to 1; cube size (CSZ), 70 to 150 mm; and curing age (CAG), 7 to 120 days. The study's compressive strength (CS), which ranged from 7.17 to 84.72 MPa, claimed different mix proportions and curing regimes. Four models were developed to predict the compressive strength (CS) of FA and SCM-based concrete using recycled aggregates, employing four different regression techniques: linear regression (LR), multiple linear regression (MLR), non-linear regression (NLR), and power equation (PE). The MLR model performed better, with a testing R2 of 0.8856 and RMSE of 4.39 MPa. The MLR model ranked second, achieving a coefficient of determination (R2) of 0.8857 and a root mean square error (RMSE) of 4.50 MPa. A sensitivity analysis revealed that W/CM and curing age were the most influential factors affecting the compressive strength. Additionally, replacing more than 40% of the coarse aggregate with recycled coarse aggregate (RCA) resulted in a 20% reduction in compressive strength, due to increased material porosity. These findings suggest that MLR-based models offer reasonably accurate predictions suitable for non-structural concrete applications, while also addressing economic and environmental considerations.

Machine-aided regression modeling for green concrete mix optimization with fly ash and recycled aggregates

期刊论文
发表日期：

2025年6月25日

查看全文

Sawan Bahaalddin Rahman,
Meer Kamil Hassan,
Ahmed Salih Mohammed

摘要

This study aims to optimize the compressive strength (CS) of eco-concrete mixes that contain fly ash (FA), recycled concrete aggregate (RCA), and supplementary cementitious materials (SCMs). Using 324 datasets, a multivariate predictive modeling framework was created to guarantee excellent quality and consistency in production processes. CS optimization is crucial as a performance measure for industrial manufacturing of these alternatives. This research aims to develop a novel combination of designs and testing methods to improve the compressive strength of sustainable building materials. By enhancing their construction sustainability, these materials can support the sector's sustainability and meet the structural needs of various applications. This study employs multivariate predictive modeling to assess important variables systematically. This approach enables a comprehensive analysis of the various factors that interact to predict outcomes. Researchers can better understand the complex relationships by considering multiple variables simultaneously. Analytical models were built using 324 datasets collected from earlier research and nine input factors, including curing age, RCA replacement ratio, and W/C ratio, to assess the performance of four regression models: power equation (PE), non-linear (NLR), linear (LR), and multilinear (MLR). The independent variables are cement, 40 to 635 kg/m³; SCMs, 0 to 592 kg/m³; W/CMs, 0.25 to 0.45; fine aggregate (FA), 676 to 1033 kg/m³; coarse aggregate (RCA), 478 to 890 kg/m³; superplasticizer (SP), 1.21 to 8.60 kg/m³; recycled content, 0 to 1; cube size (CSZ), 70 to 150 mm; and curing age (CAG), 7 to 120 days. The study's compressive strength (CS), which ranged from 7.17 to 84.72 MPa, claimed different mix proportions and curing regimes. Four models were developed to predict the compressive strength (CS) of FA and SCM-based concrete using recycled aggregates, employing four different regression techniques: linear regression (LR), multiple linear regression (MLR), non-linear regression (NLR), and power equation (PE). The MLR model performed better, with a testing R² of 0.8856 and RMSE of 4.39 MPa. The MLR model ranked second, achieving a coefficient of determination (R²) of 0.8857 and a root mean square error (RMSE) of 4.50 MPa. A sensitivity analysis revealed that W/CM and curing age were the most influential factors affecting the compressive strength. Additionally, replacing more than 40% of the coarse aggregate with recycled coarse aggregate (RCA) resulted in a 20% reduction in compressive strength, due to increased material porosity. These findings suggest that MLR-based models offer reasonably accurate predictions suitable for non-structural concrete applications, while also addressing economic and environmental considerations.