Prediction of compressive strength of concrete based on artificial neural network and sensitivity analysis of combination factors

To investigate factors affecting the compressive strength of concrete in pumped-storage facilities, this study developed a predictive analytical model based on artificial neural networks. A dataset of experimental concrete mixtures was employed to train and evaluate the model through validation and testing phases. Subsequent analysis focused on three critical parameters: water-cement ratio, cementitious material content, and water-reducing agent proportion. Validation revealed calculation errors predominantly confined to ±5%, indicating robust model reliability and strong correlation between predicted and empirical values. The analysis demonstrated an inverse relationship between concrete strength and water-cement ratio, while cementitious material content exhibited a proportional influence. Optimal compressive performance occurred at approximately 1% water-reducing agent dosage. Parameter sensitivity evaluation ranked water-cement ratio as the most significant determinant of strength variation, followed by water-reducing agent quantity, with cementitious materials showing comparatively minor effects. These findings suggest that enhanced compressive strength in pumped-storage facility concrete can be achieved through strategic formulation: minimizing water-cement ratios, increasing cementitious components, and maintaining water-reducing additives near the identified optimal threshold. The methodology provides a framework for optimizing concrete mixtures in hydroelectric infrastructure projects, balancing material efficiency with structural performance requirements.