AI-based modeling of CO2 footprint in geopolymer concrete production containing GGBFS as a by-product from the iron industry

Abstract

Abstract This study investigates the CO 2 footprint (CF) of geopolymer concrete (GPC) produced using ground granulated blast-furnace slag (GGBFS), an industrial by-product and alternative to ordinary Portland cement. Given the diversity of influential variables and the necessity for repeated testing due to changing materials and conditions, an artificial intelligence (AI)-based framework is considered an effective strategy to diminish reliance on experimental work. A novel hybrid model combining artificial neural network and biogeography-based optimization (ANN-BBO) is developed to predict the CF associated with GGBFS-based GPC production. The model incorporates 25 key variables related to precursor content and its chemical composition, activator content and characteristics, aggregate content, mix design, and curing conditions. Through rigorous evaluation using statistical metrics, error histograms, and k -fold cross-validation, the proposed ANN-BBO model demonstrated high accuracy, with 89% of its predictions falling within a 5% error margin compared to 62% for the single ANN, indicating a 27% improvement in prediction accuracy. Sensitivity analysis enhanced interpretability by identifying the superplasticizer and coarse aggregate as the key positive and negative contributors to the CF, while also suggesting that 400 kg/m 3 of GGBFS is the ideal content for maximizing compressive strength and minimizing carbon emissions in GGBFS-based GPC production.