Multi-objective optimisation for mortar containing activated waste glass powder

Abstract

Waste glass is inert and non-degradable which leads to enormous environmental and sustainability troubles, but it can be reused in concrete due to the potential of the pozzolanic activity. This study proposes methods on activity excitation of waste glass powder (WGP) including mechanical, chemical, and mechanical-chemical activation. The results showed that the mortar containing 30% 75 μm WGP activated by the mechanical-chemical method was optimal to increase the mechanical property and reduce the detrimental expansion. In addition, the microstructural analysis was conducted to explore the activation effect on WGP and WGP-cement system. An artificial intelligence (AI) based multi-objective optimisation (MOO) model was proposed to seek the optimal mix proportions for the unconfined compression strength (UCS), alkali-silica reaction (ASR), and cost. A comprehensive dataset was investigated including 549 specimens for the UCS test and 366 test results for the expansion test. Random Forest (RF) model was utilized for the prediction of UCS and ASR values with hyperparameters tuned by a firefly algorithm (FA). The high correlation coefficients (0.93 for UCS and 0.91 for ASR) verified the feasibility of FA-RF. Subsequently, the FA-RF model was extended as the objective function for the mi-objective firefly algorithm (MOFA-RF) to obtain the consequent Pareto fronts. This paper combined the results of experiments, machine learning prediction, and multi-objective optimisation design for activated WGP mortar, which provided a comprehensive basis for the practical application.

Keywords

• Waste glass powder
• Activation methodology
• Multi-objective optimisation
• Machine learning
• Unconfined compressive strength
• Alkali-silica reaction