Experimenting the compressive performance of low-carbon alkali-activated materials using advanced modeling techniques

Abstract Activated alkali materials (AAMs) are progressively utilized as an alternative to Portland cement concrete owing to their widespread application and reduced environmental impact. This research employed multi-expression programming (MEP) and gene expression programming (GEP) to create predictive models for the compressive strength (CS) of AAMs based on a dataset of 381 entries with eight distinct variables. To further assess the significance of the factors influencing the CS of AAMs, sensitivity analysis was employed. In comparison to GEP, MEP was better at predicting AAM’s CS. The R 2 score of the GEP model was 0.953, which is lower than the MEP model’s 0.970 level. This was further backed up by the results of the statistical study and Taylor’s diagram. The results of the sensitivity analysis showed that specific surface area, aggregate volumetric ratio, and silicate modulus were the three most important parameters influencing the models’ outcomes. In comparison to models built in Python, the produced models yield novel empirical equations for AAM strength characteristic prediction. Researchers and professionals in the field could use these equations to find the best proportions for mix designs, cutting down on the need for repeated laboratory tests.