Parametric analysis and prediction of geopolymerization process

Geopolymerization is a complex and dynamic process that makes its analysis complicated. The kinetic products of geopolymerization are generally investigated using calorimetric analysis. This study predicts the geopolymerization process and performs sensitivity and parametric analysis to find the influence of parameters on the geopolymerization process and suggests the threshold value for controlling the parameters for designing resilient fly ash-based geopolymers. Four parameters (geopolymerization peak heat (GPH) and peak time (GPT) and dissolution peak heat (DPH) and peak time (DPT)) were predicted using an opposition-based learning enhanced sunflower optimized boosting machine learning model. The optimized CatBoost model outperformed other models showing R² of 0.98, 0.98, 0.98, and 0.99 respectively for four output parameters. The efficacy and robustness of the developed model were verified by comparing the model with previously published models and validating the model on a new dataset. SHAP-based local sensitivity analysis resulted in curing temperature as the most critical parameter for DPH, DPT, and GPH and the liquid-to-fly ash mass ratio found critical to GPT. A parametric analysis was conducted using a 3D partial dependence plot. Optimum input content for surge and reduction of all four output parameters were deduced for designing a durable fly ash-based geopolymer. [Display omitted]