Design of solid waste-based superfine tailings cemented paste backfill based on machine learning and MAA model

作者：Yafei Hu, Ruipeng Hu, Lujing Zheng, Bin Han, Zhiyi Liu, Sitao Zhu · 发表于：Case Studies in Construction Materials · 年份：2025 · DOI：10.1016/j.cscm.2025.e05175 · 被引用次数：3 · 研究领域：Tailings Management and Properties、Rock Mechanics and Modeling、Geotechnical and Geomechanical Engineering

The application of solid waste in filling mining has become an essential direction for developing green mines. This study prepared mixed aggregates using superfine tailings (ST), fly ash (FA), and silica fume (SF), and developed a solid waste-based superfine tailings cemented paste backfill (SCPB) by using steel slag (SS), granulated blast furnace slag (GBFS), and desulfurization gypsum (FDG) as binder. The modified Andreasen-Andersen (MAA) model was used to optimize the proportion of mixed aggregates. Response surface method (RSM) experiments were conducted to investigate the development of the compressive strength of SCPB. Various microscopic testing methods were employed to reveal its hydration mechanism, and a machine learning method was used to construct an optimization model for the mix proportion of SCPB. The results indicate that the residual sum of squares (RSS) decreases with increasing ST dosage. When SF, FA, and ST are mixed in a mass ratio of 10:20:70, i.e., m(SF):m(FA):m(ST) = 10:20:70, RSS reaches a minimum RSS of 153.07, at which the mixed aggregate exhibits the lowest packing density. The compressive strength of SCPB increases with the addition of binder and slurry concentration at all curing times, and first increases then decreases with the increase of SF dosage. SS, GBFS, and FDG interact to form ettringite (AFt) and calcium silicate hydrate (C-S-H) and other hydration products, which are the fundamental source of SCPB strength. Additionally, the intellige...