An intelligent design method for new binder material activated by solid wastes by integrating attention-based tree model and heuristic optimization algorithm

Abstract

To promote the utilization of industrial solid waste, a new green binder system namely SCGF for sustainable concrete is developed, in which slag-fly ash is activated by soda residue-carbide slag. However, the mix design of the SCGF is challenging due to heavy workload and the lack of theoretical basis. The relationship between mix proportion and mechanical properties of SCGF needs to be accurately expressed. In this paper, an intelligent design method of SCGF is proposed by coupling the attention mechanism, machine learning and multi-objective heuristic optimization algorithm. After training with experimental data, the prediction performance of attention-based tree model is evaluated. Multiple Pareto optimal solutions can be obtained by the proposed intelligent mix proportion design method. The results show that attention-based tree models perform higher accuracy (accuracy > 95 %) and better measures than that of classical tree models. Compared the strength and cost of SCGF to cement mortar, the mix proportion designed in this study has obvious advantages in environmental protection and economy (>50 % cost saving). Therefore, the proposed method can provide a theoretical basis for the recycling and utilization of industrial solid wastes.