Scholarlink

<p>Ultra-fine tailings cemented paste backfill (UCPB) is increasingly adopted in mining operations due to its efficiency and environmental benefits. However, the thixotropic rheological behavior of UCPB during pipeline transportation remains a critical factor affecting the technology. This study investigates the thixotropic properties of UCPB and develops a stacking machine learning model to predict its initial and equilibrium rheological parameters accurately. Constant shear thixotropy tests were conducted under varying mass concentrations, cement-to-tailings ratios (CTR), and particle gradations characterized by packing density. The results indicate that UCPB exhibits shear-thinning behavior, with shear stress decreasing and stabilizing over time. Mass concentration showed a positive correlation with rheological parameters, while CTR and packing density were negatively correlated; the feature importance ranking is mass concentration &gt; packing density &gt; CTR. The proposed stacking model, which integrates Bayesian-optimized extreme gradient boosting (BO-XGBoost) and backpropagation neural network (BO-BPNN) as base models, demonstrated superior predictive accuracy over single models, maintaining relative errors around 5% and showing significant improvements in <i>R</i>-squared, RMSE, and MAPE metrics. Industrial tests validated the thixotropic characteristics of UCPB and the model’s effectiveness; pipeline resistance decreased with flow time, eventually stabilizing, and the resistance values calculated using the model’s predicted thixotropic rheological parameters closely matched the monitoring values. The findings confirm that the proposed stacking model is an efficient and accurate tool for predicting the thixotropic rheological parameters of UCPB, facilitating the effective utilization of ultra-fine tailings as filling material in mining engineering.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Intelligent Design Framework for Predicting Thixotropic Rheological Parameters of Ultra-Fine Tailings Cemented Paste Backfill: A Stacking Machine Learning Model

  • 期刊论文

  • 发表日期：

    2025年2月10日

查看全文
  • Jinlong Yao,
  • Tianyu Yang,
  • Dengpan Qiao,
  • Haiyong Cheng,
  • Hao Chen

摘要

Ultra-fine tailings cemented paste backfill (UCPB) is increasingly adopted in mining operations due to its efficiency and environmental benefits. However, the thixotropic rheological behavior of UCPB during pipeline transportation remains a critical factor affecting the technology. This study investigates the thixotropic properties of UCPB and develops a stacking machine learning model to predict its initial and equilibrium rheological parameters accurately. Constant shear thixotropy tests were conducted under varying mass concentrations, cement-to-tailings ratios (CTR), and particle gradations characterized by packing density. The results indicate that UCPB exhibits shear-thinning behavior, with shear stress decreasing and stabilizing over time. Mass concentration showed a positive correlation with rheological parameters, while CTR and packing density were negatively correlated; the feature importance ranking is mass concentration > packing density > CTR. The proposed stacking model, which integrates Bayesian-optimized extreme gradient boosting (BO-XGBoost) and backpropagation neural network (BO-BPNN) as base models, demonstrated superior predictive accuracy over single models, maintaining relative errors around 5% and showing significant improvements in R-squared, RMSE, and MAPE metrics. Industrial tests validated the thixotropic characteristics of UCPB and the model’s effectiveness; pipeline resistance decreased with flow time, eventually stabilizing, and the resistance values calculated using the model’s predicted thixotropic rheological parameters closely matched the monitoring values. The findings confirm that the proposed stacking model is an efficient and accurate tool for predicting the thixotropic rheological parameters of UCPB, facilitating the effective utilization of ultra-fine tailings as filling material in mining engineering.