Intelligent optimization for modeling carbon dioxide footprint in fly ash geopolymer concrete: A novel approach for minimizing CO2 emissions

Abstract

In the global effort to mitigate climate change and reduce CO2 emissions, this study introduces an innovative, pioneering approach that combines artificial intelligence and experimental methods to investigate the CO2 footprint (CO2-FP) in fly ash geopolymer concrete materials. Three powerful non-linear intelligent learners, including Gaussian Process Regression (GPR) with Response Surface Methodology (RSM), Support Vector Regression (SVR), and Standalone Decision Tree Regression (DTR) are employed. The models are developed using seven input features related to the curing temperature, fly ash content, concentrations of coarse and fine aggregates, alkaline activators (Na2SiO3, NaOH) content, and superplasticizer. To identify the most influential input features, three different combinations (combo-1, combo-2, and combo-3) of these features are utilized in model building. The models' performance is assessed using key metrics such as coefficient of correlation (CC), Nash Sutcliffe coefficient efficiency (NSE), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). During the verification phase, the GPR-3 [Combo-3] model emerges as the most efficient in predicting the CO2-FP, with a high CC value of 0.9645 and NSE value of 0.9292. Consistently, Combo-3 demonstrates superior performance across all the models, underscoring the significance of the selected features. The findings of this study provide valuable guidance to industries and policymakers, enabling them to optimize concrete compositions and minimize CO2 emissions, thus contributing to global environmental sustainability.