<![CDATA[The growing demand for sustainable yet high-performance construction materials has intensified research into alternatives to Ordinary Portland Cement (OPC), whose production accounts for approximately 7–8% of global CO₂ emissions. Geopolymer Concrete (GPC), synthesized through the alkali activation of aluminosilicate-rich industrial by-products, has emerged as a promising low-carbon binder. However, the design of Ultra-High-Performance Geopolymer Concrete (UHGC), typically characterized by compressive strengths exceeding 120 MPa, remains highly complex due to the strong sensitivity of mechanical performance to mix composition, activator chemistry, and reinforcement parameters. This study proposes a transparent, data-driven framework for predicting and optimizing UHGC compressive strength using Multiple Linear Regression (MLR). A comprehensive dataset comprising 72 UHGC mixtures (122.9–168.8 MPa) was compiled, incorporating key variables including precursor ratio, Si/Al ratio, steel fiber volume fraction, superplasticizer content, and water-to-binder ratio. The MLR model demonstrated excellent predictive accuracy and generalization, achieving R² values of 0.944 and 0.921 for training and testing datasets, respectively, with low RMSE (~4.5 MPa). Statistical analysis confirmed the dominance of the Si/Al ratio and water-to-binder ratio as the most influential parameters governing UHGC strength. Experimental validation using nine independently designed UHGC mixtures further confirmed the robustness of the model, yielding a high correlation between predicted and measured strengths (R² = 0.954) with a mean absolute percentage error below 1%. The optimal formulation achieved a compressive strength of 168.8 MPa at a Si/Al ratio of approximately 6.0 with 1.0% steel fiber content. Compared to more complex machine learning models, the proposed MLR approach offers competitive accuracy while retaining full interpretability, enabling rational mix design and informed decision-making. This study demonstrates that interpretable predictive modeling can effectively bridge geopolymer chemistry and UHGC mix optimization, providing a practical and sustainable pathway for the development of next-generation ultra-high-performance construction materials.]]>
Copyrights © 2026
