Abstract

Geopolymer concrete (GC) is a cleaner construction material that has the potential to replace conventional Portland cement concrete (PCC) completely. The manufacturing process of geopolymer concrete results in a reduced carbon footprint and consumes less energy compared to that of Portland cement. With this motivation, this work develops a physics-informed machine learning model to predict the compressive strength of the geopolymer concrete. The method leverages experimental datasets and exploits knowledge of concrete chemo-mechanical governing equations to identify variations in the compressive strength under the application of fly ash. Model predictions align with the experimental findings with an R-Squared of 0.933 and a RMSE of 3.35 MPa. The key innovation of this work is the aggregation of physical information and data-driven models to provide a robust, automated simulation platform for geopolymer concrete. The developed model will help in decision-making and guide the use of geopolymer concrete as an alternative to Portland cement. This will contribute to cleaner production because of the significant carbon dioxide reductions while promoting geopolymer concrete in construction.

Keywords: Geopolymer concrete; fly ash; machine learning; compressive strength; prediction