Performance evaluation of geopolymer masonry units: A hybrid approach combining laboratory testing and AI modeling

Abstract

This study presents a comprehensive investigation into the compressive strength and stress–strain behavior of geopolymer brick masonry, focusing on both prisms and wallettes. Geopolymer bricks and mortars were used to fabricate specimens, and their mechanical performance was experimentally evaluated. The study also employs nine machine learning algorithms on a dataset comprising 612 prism and 63 wallette data points, assessing performance based on six predictive metrics. Experimental results revealed that prisms exhibited higher compressive strength (7.2 MPa to 2.6 MPa) compared to wallettes (6.5 MPa to 1.2 MPa), with a linear regression indicating wallettes achieve approximately 88 % of prism strength. Among the ML models, Random Forest performed best, with R² values of 0.92 and 0.97 for prism and wallette datasets, respectively. The results emphasize the influence of brick-and-mortar properties and dimensional parameters on masonry performance. This study advances the understanding of geopolymer masonry and demonstrates the synergy of experimental analysis and machine learning for predictive modeling in sustainable construction.