Data-driven rheological model for 3D printable concrete

Additive manufacturing in construction demands an in-depth understanding of the rheological properties of fresh concrete. However, the abundant data in this field remains underexplored. This conventional fragmented approach has hindered broader progress and innovation. This study aims to develop rheological models for 3D printable concrete through a comprehensive, data-driven paradigm, emphasizing the urgent need for a unified, large-scale dataset. By compiling data spanning a decade, we have created an open-access dataset that contains mix designs and experimental results on the rheological behaviors of additive construction concrete. A machine learning-based model and explicit polynomial expressions for estimating rheological properties were developed. The developed machine learning model can take nineteen different parameters as inputs to predict the rheological behavior of printed concrete, showing superiority over models considering a few parameters. Our model can predict the properties of unexplored mix designs, with tailored expressions for practical engineering in additive construction. This enhances understanding of concrete mix design and rheology, highlighting the importance of data-driven method in unveiling the complexity of concrete. •Analyzed the rheology of 3D printing concrete through the lens of data.•Identified research gaps in the area of mix designs for 3D printing concrete.•Developed novel predictive models for 3D printing concrete and explained the models using game theory.•Developed novel explicit equations for rapid estimation of rheological properties based only on the mix design.