Optimization of cleaner masonry blocks: Response surface methodology and genetic algorithm integrated neural networks in the applications of slag and waste sand

Maximizing the reuse of industrial waste while producing high-performance building materials is essential for advancing a circular economy and promoting material sustainability. In this context, this study investigates the mechanical performance and optimization of masonry blocks produced from waste foundry sand (WFS) and ground blast furnace slag (GBFS) through alkali-activation. The effects of curing temperature, binder content, and water-to-solid ratio on the unconfined compressive strength (UCS) were systematically evaluated. The highest UCS of 16.7 MPa was obtained at 80 °C with 50 % binder content and a water-to-solid ratio of 0.15, supported by dense microstructures and effective calcium-silicate-hydrate (C-S-H) gel formation, confirmed by X-ray diffraction (XRD). UCS increased by 59 % when curing temperature rose from 40 °C to 80 °C but decreased by 36 % at 100 °C due to moisture loss and zeolitic phase formation. Higher binder content improved interfacial transition zones, whereas lower content (20 %) resulted in unreacted particles and reduced UCS. While Response Surface Methodology (RSM) captured linear and quadratic interactions, several terms had p-values >0.05, which could lead to overfitting. Therefore, a feedforward neural network (ANN) with mean square error <10 % was employed to capture complex nonlinear interactions and guide Genetic Algorithm (GA)-based optimization, achieving predicted UCS values within 5 % relative error. Overall, this study demonstrates that GA-ANN modeling enables the efficient valorization of industrial wastes into high-strength masonry blocks, advancing material sustainability, circular economy practices, and resilient construction solutions.