Development of a sustainable quaternary blended geopolymer matrix: optimization of waste glass powder utilization through integrated Taguchi and neural network approaches

This investigation evaluates the important parameters affecting the fresh and hardened properties of a quaternary blended ground granulated blast furnace slag (GGBS)-based geopolymer matrix optimized through the Taguchi approach. Four variables were investigated using an L16 orthogonal array: waste glass powder (WGP), metakaolin (MK), pond ash (PA) and sodium hydroxide (NaOH) molarity, each at four levels. The ANOVA analysis showed that the content of WGP and SH molarity are the most significant, accounting for 86.70% and 6.57% of the variance in workability and 48.63% and 21.23% of the variance in compressive strength, respectively. The mix proportion comprising 40% WGP, 5% MK, 10% PA, and 12 M of NaOH solution showed better mechanical properties, with a compressive strength of 61.84 MPa, split tensile strength of 5.93 MPa, and flexural strength of 7.35 MPa, respectively, among the 16 mixes. Signal-to-noise ratio (SNR) analysis showed that the enhanced mix produced compressive, split tensile and flexural strengths of 65.35 MPa, 6.12 MPa and 7.83 MPa, respectively. All the mechanical and workability properties were predicted by an ANN model, with prediction errors less than 10% and R values above 0.95, verifying the reliability of the proposed approach.