An-Najah University Journal for Research - A (Natural Sciences)

Article info

In Press

Keywords

• Machine Learning.
• Sustainable Construction
• Self-Healing Concrete
• Crack Prediction
• Evolutionary Algorithms

Abstract

Self-healing concrete (SHC) represents a sustainable innovation capable of autonomously repairing cracks, thereby reducing maintenance costs, energy consumption, and environmental impact. However, optimizing SHC design remains a challenge due to the nonlinear interactions among material constituents and healing mechanisms. This study investigates the predictive capability of two evolutionary algorithms Gene Expression Programming (GEP) and Multi-Expression Programming (MEP) to estimate the cracked area (CrA) in SHC mixtures incorporating polymer fibers and alkali-resistant bacteria. A dataset of 1007 records defined by six variables (cement, fine aggregate, water, curing time, bacterial concentration, and fiber dosage) was analyzed to establish interpretable mathematical models. Model validation demonstrated that both frameworks effectively reproduced experimental results, though the MEP model exhibited superior performance with R² = 0.991, RMSE = 0.150 mm², and MAE = 0.120 mm², compared with the GEP model (R² = 0.975, RMSE = 0.322 mm², MAE = 0.262 mm²). Broader statistical indicators, including NSE, RRMSE, and a₂₀ index (0.950 for MEP versus 0.830 for GEP), confirmed the higher accuracy, consistency, and robustness of MEP. Furthermore, SHapley Additive exPlanations (SHAP) analysis highlighted curing time and bacterial concentration as the most influential variables, while fiber showed a nonlinear dual effect depending on dosage. The developed formulations not only enhance predictive accuracy but also provide interpretable insights for optimizing SHC mix design, contributing to the advancement of durable and sustainable construction materials.

معلومات المقال

قيد النشر

الكلمات الإفتتاحية

• Machine Learning.
• Sustainable Construction
• Self-Healing Concrete
• Crack Prediction
• Evolutionary Algorithms

الملخص

Self-healing concrete (SHC) represents a sustainable innovation capable of autonomously repairing cracks, thereby reducing maintenance costs, energy consumption, and environmental impact. However, optimizing SHC design remains a challenge due to the nonlinear interactions among material constituents and healing mechanisms. This study investigates the predictive capability of two evolutionary algorithms Gene Expression Programming (GEP) and Multi-Expression Programming (MEP) to estimate the cracked area (CrA) in SHC mixtures incorporating polymer fibers and alkali-resistant bacteria. A dataset of 1007 records defined by six variables (cement, fine aggregate, water, curing time, bacterial concentration, and fiber dosage) was analyzed to establish interpretable mathematical models. Model validation demonstrated that both frameworks effectively reproduced experimental results, though the MEP model exhibited superior performance with R² = 0.991, RMSE = 0.150 mm², and MAE = 0.120 mm², compared with the GEP model (R² = 0.975, RMSE = 0.322 mm², MAE = 0.262 mm²). Broader statistical indicators, including NSE, RRMSE, and a₂₀ index (0.950 for MEP versus 0.830 for GEP), confirmed the higher accuracy, consistency, and robustness of MEP. Furthermore, SHapley Additive exPlanations (SHAP) analysis highlighted curing time and bacterial concentration as the most influential variables, while fiber showed a nonlinear dual effect depending on dosage. The developed formulations not only enhance predictive accuracy but also provide interpretable insights for optimizing SHC mix design, contributing to the advancement of durable and sustainable construction materials.