Scholarlink

<p>In the field of mining industry, drilling performance is a critical factor influencing the efficiency and cost-effectiveness of mining operations. Accurate prediction of the rate of penetration (ROP) in rotary drilling is essential for optimizing efficiency, reducing costs, and improving safety in surface mining operations. In this study, six machine learning algorithms, including XGBoost, were trained and evaluated using rock mechanics and operational data from three Iranian surface mines, and the top-performing model was further enhanced with jellyfish search optimization (JSO) to predict ROP more accurately using the collected 73 data samples from three different mines in Iran. The results demonstrate that the XGBoost-JSO hybrid model outperforms the other models, achieving an <i>R</i><sup>2</sup> value of 0.99988 and an RMSE of 0.19999 in the training phase, and an <i>R</i><sup>2</sup> of 0.99336 and an RMSE of 1.16902 in the test phase. The gray wolf optimizer (GWO) and ant lion optimizer (ALO) optimized models also significantly improved over the baseline XGBoost model, with reduced error metrics and enhanced predictive accuracy. In addition, a sensitivity analysis was conducted to assess the relative importance of the input parameters (UCS, SH, RPM, Thrust, and F<sub>RMR</sub>) influencing ROP predictions. The results revealed that UCS is the most influential parameter with a score of 0.985, followed by RPM (0.917) and SH (0.855).</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

Prediction of Rotary Drilling Rate of Penetration in Surface Mines Using Machine Learning Techniques

  • 期刊论文

  • 发表日期：

    2025年11月3日

查看全文
  • Zahra Nabavi,
  • Shahab Hosseini,
  • Jafar Khademi Hamidi,
  • Masoud Monjezi,
  • Fakhrossadat Hasheminasab,
  • Ali Nemati Vardin,
  • Mahdi Hasanipanah

摘要

In the field of mining industry, drilling performance is a critical factor influencing the efficiency and cost-effectiveness of mining operations. Accurate prediction of the rate of penetration (ROP) in rotary drilling is essential for optimizing efficiency, reducing costs, and improving safety in surface mining operations. In this study, six machine learning algorithms, including XGBoost, were trained and evaluated using rock mechanics and operational data from three Iranian surface mines, and the top-performing model was further enhanced with jellyfish search optimization (JSO) to predict ROP more accurately using the collected 73 data samples from three different mines in Iran. The results demonstrate that the XGBoost-JSO hybrid model outperforms the other models, achieving an R2 value of 0.99988 and an RMSE of 0.19999 in the training phase, and an R2 of 0.99336 and an RMSE of 1.16902 in the test phase. The gray wolf optimizer (GWO) and ant lion optimizer (ALO) optimized models also significantly improved over the baseline XGBoost model, with reduced error metrics and enhanced predictive accuracy. In addition, a sensitivity analysis was conducted to assess the relative importance of the input parameters (UCS, SH, RPM, Thrust, and FRMR) influencing ROP predictions. The results revealed that UCS is the most influential parameter with a score of 0.985, followed by RPM (0.917) and SH (0.855).