Development of heat-resistant tunnel muck-based shotcrete for geothermal environments: Dual drive of combining explainable machine learning and microstructure characterization

Abstract

The in-situ material shortage and high geothermal service conditions significantly hinder the utilization of shotcrete for the construction of tunnels in plateau regions. In this study, fully tunnel muck-based shotcrete (FTMS) was developed using stone powder and sand-gravel aggregates derived from tunnel muck. The effects of stone powder content, stone powder modification methods, and six performance-reinforcing materials on the high-geothermal resistance of FTMS were systematically investigated. The results demonstrated that a stone powder content of 20 % and a modification ratio of stone powder:silica fume:modifier = 82:15:3 achieved an optimal balance between stone powder utilization and high-geothermal performance. The improvement effect of calcium carbonate whiskers on the compressive strength of FTMS under high geothermal conditions was most significant, while the optimization effect of nano alumina sol on compressive strength and mass loss rate was most prominent. The influence of dual-doped performance-reinforcing materials on three properties was predicted using artificial neural networks, and interpretability analysis was conducted using SHAP. Finally, the optimal combination strategy was determined to be the co-doping of nano alumina sol and calcium carbonate whiskers. Microstructural analysis revealed that the addition of nanomaterials forms a dense calcium carbonate network in FTMS, while promoting the secondary hydration of calcium hydroxide, resulting in high-strength hydration products such as monocarboaluminate and ettringite. Additionally, a significant transformation of C-S-H to C-A-S-H phases was observed, accompanied by a 36.6 % reduction in harmful pores.