Investigating the influence of mix design parameters on compressive strength in preplaced-aggregate green concrete using predictive models

This study presents a predictive framework for estimating the compressive strength of preplaced aggregate concrete (PAC) using a comprehensive dataset and advanced statistical modeling. A total of 261 concrete mix samples were compiled, each incorporating various combinations of materials such as cement, fly ash, silica fume, GGBS, sand, gravel, water, superplasticizer, and expanding admixtures. Key mix design parameters like the water-to-binder (W/B) and sand-to-binder (S/B) ratios were systematically varied to reflect realistic construction practices. To identify the most influential components and improve model performance, data normalization and sensitivity analysis were performed. The analysis revealed that the W/B ratio was the most critical factor, contributing approximately 31.5% to compressive strength variation. The independent variable ranges in the dataset are as follows: cement (176–873 kg/m3), fly ash (0–262 kg/m3), silica fume (0–57 kg/m3), GGBS (0–228 kg/m3), sand (0–873 kg/m3), water (100–431 kg/m3), gravel (1.5–2001 kg/m3), water to cement ration (W/B) ranged between 0.3–0.85, S/B (0–2), superplasticizer (0–10.9 kg/m3), and expanding admixture (0–58.6 kg/m3). Compressive strength, the dependent variable, ranged from 5.7 MPa to 58.6 MPa. Sensitivity analysis identified W/B as the most influential variable, showing a sensitivity of 31.5% across samples. After testing multiple models, the Full Quadratic (FQ) model emerged as the most accurate based on RMSE, MAE, and OBJ performance criteria. The strength values ranged from 5.7 MPa to 58.6 MPa, encompassing low- to high-strength concrete applications. Among several tested models, the Full Quadratic (FQ) model demonstrated the highest prediction accuracy based on key evaluation metrics (RMSE, MAE, and objective function). This model offers a reliable tool for engineers to estimate compressive strength and optimize mix design without extensive laboratory testing. The proposed approach contributes to reducing construction costs, enhancing design efficiency, and supporting data-driven decision-making in sustainable concrete development.