Modeling and optimization of Multi-Waste foam geopolymers using central composite design and neural networks.

The development of sustainable construction materials with improved mechanical performance and thermal insulation and is crucial to addressing growing environmental and energy efficiency concerns. This study aimed to design and optimize foamed geopolymer composites incorporating recycled concrete sand (RCS), glass powder (GP), and date palm fibers (DPF) to improve compressive strength and thermal insulation. A comparative analysis of two modeling strategies, Central Composite Design (CCD) and Artificial Neural Network (ANN), was conducted to predict and optimize the effects of the three constituents. Experimental data were used to train and validate both models. The ANN model outperformed CCD in terms of predictive accuracy, particularly for thermal conductivity, as indicated by higher R² values and a greater overall desirability index (0.86 vs. 0.77). The ANN-optimized formulation (35.4% RCS, 4.96% GP, 0.23% DPF) achieved a compressive strength of 4.41 MPa and thermal conductivity of 0.141 W/mK. While ANN offers improved predictive capability, its limited interpretability remains a challenge. This research highlights the potential of integrating construction and agricultural waste into geopolymer matrices to produce eco-efficient building materials.