Data driven predictive analysis of hardened properties of sustainable geopolymer concrete developed by using fly ash and sugarcane bagasse ash using explainable AI

This study investigates the mechanical performance, predictive modelling, and sustainability characteristics of geopolymer concrete developed using FA and SBA as binary precursor materials. Experimental investigations were conducted to evaluate Cs, Sts, and Fs under varying SBA replacement levels (0–20%) and ambient curing conditions. The results indicate that although increasing SBA content leads to a gradual reduction in strength, mixes containing 10–15% SBA exhibit an optimal balance between mechanical performance and material sustainability due to improved geopolymer gel formation and matrix densification. To capture the intricate nonlinear relationships that control strength behaviour, machine learning models such as the Random Forest, Gradient Boosting, Extreme Gradient Boosting and a hybrid model were used. The hybrid model was the best predictor with the best generalization to all the strength responses. The most significant parameters were found to be curing age and SBA content, which was in line with the trends in the experiment explained by AI analysis. The interactive condition of mix composition and curing conditions was also demonstrated by three-dimensional response surface analysis. Sustainability assessment indicated that FA–SBA based BGPC reduces embodied energy and carbon emissions compared to conventional Ordinary portland cement-based concrete. Although alkaline activators increase initial production cost, large-scale implementation can improve economic feasibility. The integrated experimental–ML–sustainability framework confirms FA–SBA Geopolymer concrete as a reliable and sustainable environmentally alternative construction material.