A multi-objective optimisation approach for activity excitation of waste glass mortar

Waste glass is promising to be recycled and reused in construction for sustainability. Silicon dioxide is the main component of glass, however, its pozzolanic activity is latent mainly due to its stable silica tetrahedron structure. To excite the activation of waste glass, chemical activation and mechanical grinding of waste glass powder (WGP) were investigated. As the supplementary, hydrothermal and combined (mechanical-chemical-hydrothermal) treatments were conducted on part of the WGP samples. The unconfined compression strength (UCS), expansion caused by alkali–silica reaction (ASR), and the microstructural morphology of WGP were investigated. The results showed the dosage threshold (around 2%) of the chemical activators (alkali and sodium sulfate) and the combined activation were optimal. Besides, a firefly algorithm (FA) based multi-objective optimisation model (MOFA) was applied to seek the Pareto fronts based on three objectives: UCS, ASR expansion, and Cost of mixture proportion. The objective functions of UCS and expansion were established through training the machine learning (ML) models where FA was used to tune the hyperparameters. The cost was calculated by a polynomial function. The ultimate values of root mean square error (RMSE) and correlation coefficient (R) showed the robustness of the ML models. Moreover, the Pareto fronts for mortars containing 300 μm and 75 μm WGPs were successfully obtained, which contributed to the practical application of waste glass in mortar production. In addition, the sensitivity analysis was conducted to rank the importance of input variables. The results showed that curing time, activator's content, and WGP particle size were three essential parameters.