Recycling three local waste materials for optimizing the mechanical performance of fly ash-based geopolymer cement: Application of a ternary mixture design and artificial neural networks modeling

Geopolymer technology offers a promising approach for recycling inorganic solid waste and enhancing the properties of sustainable building materials. This research combines eco-friendly waste recycling with advanced statistical and machine learning techniques to improve the mechanical performance of fly ash-based geopolymer cement. Ternary Mixture Design (TMD) in conjunction with Artificial Neural Networks (ANN) is employed to optimize key characteristics, including compressive strength, bulk density, and apparent porosity. The study utilized three types of local waste—ceramic waste (CW), coal bottom ash (CBA), and marble powder (MP)—as an inorganic reinforcement system. The optimization through TMD led to an ideal combination of (40 % CW-41 % CBA-19 % MP), resulting in a high-performing geopolymer with a compressive strength of 27.63 MPa, an apparent density of 1.93 g/cm³, and an apparent porosity of 22.07 %. Both TMD and ANNs accurately predicted the experimental results, with ANNs exhibiting slightly greater precision. Physicochemical characterizations confirmed that the optimized geopolymer (OGP) displayed a compact and homogeneous matrix dominated by an amorphous sodium-calcium aluminosilicate hydrate (N(C)-A-S-H) gel phase. In contrast, the non-performing geopolymer (NPGP) system revealed a porous and heterogeneous structure. This work demonstrates that integrating waste reuse with advanced modeling tools can produce high-performance and sustainable geopolymer materials. The findings advocate for the development of scalable, eco-friendly alternatives for the construction industry and emphasize the benefits of merging traditional statistical methods with machine learning in materials engineering.