A novel hybrid metaheuristic algorithm and response surface methodology approach for predictive modeling and optimization of cementitious compressive strength

Compressive strength is a fundamental property for evaluating cementitious materials, yet its prediction and optimization remain challenging due to the complex and nonlinear interactions of chemical composition, curing time, and processing conditions. This study presents a hybrid framework that integrates Response Surface Methodology (RSM) with the Genetic Algorithm (GA) to enhance predictive modeling and optimize cement mixtures. RSM was used to construct polynomial models-linear, linear + square, linear + interaction, and full quadratic-to capture the relationships between eight oxide compositions (Silo₂, Allow₃, Foo₃, Cano, Mg, SO₃, K₂O, NaoSiO₂,Al₂O₃,Fe₂O₃,CaO,MgO,SO₃ ,K₂O,Na₂O) and compressive strength at 2, 7, and 28 days. The Genetic Algorithm was employed to explore this high-dimensional solution space and identify optimal compositions. Experimental data from a regional cement producer /laboratory were used for model training and validation. Results indicate that the full Quadratic model consistently outperformed other RSM models, achieving the lowest RMSE and highest R² across all curing periods. GA optimization further improved performance, with maximum predicted compressive strengths increasing by 83% at 2 days, 48% at 7 days, and 43% at 28 days compared to RSM-only predictions. These findings highlight the robustness of the hybrid GA + RSM approach in capturing nonlinearities, improving accuracy, and achieving superior optimization outcomes. The framework offers a promising tool for developing eco-efficient and high-strength cementitious mixtures, contributing to innovation in sustainable construction materials.