Compressive strength prediction of high-performance concrete: Integrating multi-ingredient influences and mix proportion insights

The mechanical properties of high-performance concrete (HPC) are determined by the type, properties, and proportions of its material ingredients. Considering the flexibility of the mix proportion for the various ingredients and the highly nonlinear relationships among variables, empirical models cannot reliably predict the mechanical properties of HPC. This study introduces a Light Gradient Boosting Machine (Light-GBM) model optimized with Optuna tuning, which effectively predicts the compressive strength of HPC across varying mix ratios and identifies the critical parameters influencing prediction accuracy. The content of water, cement, coarse aggregate, and fine aggregate serve as pivotal factors in determining the compressive strength of concrete. The contribution of superplasticizer, blast furnace slag, and fly ash to the strength of concrete is somewhat less pronounced. The exploration of optimal mix proportions using genetic algorithms have revealed that these mixes typically achieve compressive strengths from 70 to 80 MPa. Statistical analysis of these optimal mixes underscores that water-cement ratios mainly range from 0.25 to 0.4, coarse to fine aggregate ratios are approximately 1.5, blast furnace slag ratios extend from 0.8 to 1.6, and superplasticizer to cement ratios stabilize around 0.02. Finally, parametric studies are conducted for material constituents considering their contribution to predictions. The developed prediction models demonstrate high accuracy and generalization, revealing the complex interactions among feature variables in determining the mechanical properties of HPC and providing crucial insights into optimizing HPC mixes.