Mix design of sustainable concrete using generative models

Concrete is facing scrutiny due to its substantial carbon footprint. A sustainable transformation in concrete production becomes paramount. Traditional methods for creating greener concrete mixes heavily rely on empirical, iterative testing, which is slow, expensive, and often suboptimal. This study introduces a new optimization framework that leverages the capabilities of a generative model to transform the intricate, high-dimensional space of concrete mix designs into a more manageable, two-dimensional latent space. By performing optimization in this latent space, the method efficiently approximates optimal concrete mix designs that are not only environmentally friendly but also economically viable. Moreover, utilizing an explainable regression model and analyzing the distribution of generated mix designs, the method ensures that machine generated mix designs are realistic and practically feasible. The optimization concurrently addresses multiple objectives such as compressive strength, CO2 emissions, and cost. Testing on a publicly available dataset, the results validate the efficacy of the proposed framework in optimizing concrete mix designs, yielding both sustainable and cost-effective solutions, indicating the potential for widespread adoption in the construction industry. •An efficient generative framework for sustainable concrete mix design is proposed.•The framework reduces computational difficulty by shrinking the design space dimension.•The framework generates realistic mix designs that respect ingredient interactions.•The models used in the framework are explainable.