Kai Yang1,Mathieu Bauchy1
University of California, Los Angeles1
Kai Yang1,Mathieu Bauchy1
University of California, Los Angeles1
Concrete—which is by far the most manufactured material in the world—is responsible for 5-to-10% of human CO<sub>2</sub> emissions. Here, we present an uncertainty-aware, machine-learning-enabled optimization scheme that aims to accelerate the discovery of new optimized concrete mixes featuring minimum embodied CO<sub>2</sub> while meeting target performance and manufacturing constraints. We curate an unprecedented dataset comprising more than 1 million concrete mixtures with varying mixture proportions, together with their measured properties (compressive strength, slump, shrinkage, setting time, etc.). The dataset is used to train a series of Gaussian Process regression (GPR) forward models that accurately map concretes’ mixture proportions to their properties, and uncertainty thereof. We then introduce a new inverse design approach that simultaneously leverages (i) multi-property predictions from the GPR model, (ii) uncertainty thereof, and (iii) physical knowledge to guide the discovery of new concrete mixtures featuring minimum embodied CO<sub>2</sub> while presenting required performance metrics (e.g., with a strength meeting or exceeding a given target) and obeying manufacturing constraints (e.g., with compliant slump, pumpability, finishability, etc.). This pipeline leads to the discovery of several new concrete mixtures presenting a >50% decrease in embodied CO<sub>2</sub>, with no cost increase.