EcoBlendNet: A physics-informed neural network for optimizing supplementary material replacement to reduce the carbon footprint during cement hydration

Abstract

The addition of supplementary cementitious materials (SCMs) to cement triggers a complex chemo-physics from intricate mineral admixture interactions. This work develops EcoBlendNet, a novel physics-informed neural network (PINN), to analyze carbon emissions during SCMs-enhanced cement hydration. EcoBlendNet integrates experimental data and the chemo-physical aspects of cement hydration in a heated cement paste for various mixing blends, including Portland cement, cement-fly ash blends, and cement-slag blends. EcoBlendNet accurately predicts concrete maturity and strength, capturing early-age temperature rises. Achieving a remarkable relative L2 error of 0.03 with just 5% of training data, it facilitates efficient mesh-free analysis. The study demonstrates that transitioning from cement-blend to cement-fly ash and cement-slag blends can significantly reduce CO2 emissions without compromising strength development. Quantitative analysis suggests that replacing 45–80% of cement with industrial fly ash and slag can decrease CO2 emissions by 60–80% during cement hydration. Thus, EcoBlendNet, validated through experimentation, offers a practical and eco-friendly approach to concrete construction, bridging the gap between theory and practice.