Closed-loop optimization using machine learning for the accelerated design of sustainable cements incorporating algal biomatter

Abstract

The substantial embodied carbon of cement, coupled with the ever-increasing need for construction materials, motivates the need for more sustainable cementitious materials. An emerging strategy to mitigate CO₂ emissions involves incorporating carbon-negative biomatter; however, this introduces new challenges due to complex hydration-strength relationships and the combinatorial design space. Here, using machine learning, we develop a closed-loop optimization strategy to accelerate green-cement design with minimal CO₂ emissions while meeting compressive-strength criterion. Green cements incorporating algae are tested in real time to predict strength evolution, with early-stopping criteria applied to accelerate the optimization process. This approach, using only 28 days of experiment time, attains both the strength requirement and 93% of the achievable improvement in global warming potential (GWP), resulting in a cement that has a 21% reduction in GWP. We further validate model-informed relationships via analysis of hydration, demonstrating the potential for developing materials grounded in scientific understanding.