Scholarlink

<p>Water inrush from the coal seam floor is a nonlinear phenomenon driven by the interaction of hydrogeological and mining factors. Microseismic (MS) and groundwater systems can sensitively capture fracture propagation and seepage characteristics. Based on this, this study has conducted the characterization of fracture-dense areas, correlation analysis of precursor indicators, and the development of an intelligent early warning model for two typical mines. First, a five-dimensional MS clustering degree (MSCD) integration system is developed, incorporating time, spatial three-dimensional, and energy. The three-dimensional kernel density estimation (3D-KDE) method based on MSCD is employed to reveal the multi-dimensional clustering distribution characteristics of floor instability and rupture. Through field monitoring and analysis, the evolutionary sequence of water inrush characterized by “intense mining disturbance–high KDE–water level decline–high water inrush” is identified. Given the complexity and discreteness of the quantitative indicator data, a long short-term memory (LSTM)–multi-head attention (MHA)–Bayesian optimization (BO) model is developed. Evaluation results indicate that the model demonstrates accuracy, precision, recall, and F<sub>1</sub> score of 0.8901, 0.8476, 0.8476, and 0.8436, respectively, outperforming traditional machine learning and deep learning models. Field applications further show that the LSTM–MHA–BO model achieves coefficient of determination (R<sup>2</sup>), root mean square error (RMSE), and mean absolute error (MAE) values of 0.924/0.971, 0.06/0.04, and 0.04/0.04, respectively, and that it successfully provides early warnings for high water inrush events in both mines.</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

Quantitative Early Warning of Floor Water Inrush Based on Deep Learning: Joint Response of Microseismic Three-Dimensional Kernel Density and Hydrodynamics

  • 期刊论文

  • 发表日期：

    2025年11月27日

查看全文
  • Hang Li,
  • Xianjie Hao,
  • Jinyi Ji,
  • Huaixiang Yang,
  • Wantong Lin,
  • Sheng Zhang,
  • Xinlei Cui

摘要

Water inrush from the coal seam floor is a nonlinear phenomenon driven by the interaction of hydrogeological and mining factors. Microseismic (MS) and groundwater systems can sensitively capture fracture propagation and seepage characteristics. Based on this, this study has conducted the characterization of fracture-dense areas, correlation analysis of precursor indicators, and the development of an intelligent early warning model for two typical mines. First, a five-dimensional MS clustering degree (MSCD) integration system is developed, incorporating time, spatial three-dimensional, and energy. The three-dimensional kernel density estimation (3D-KDE) method based on MSCD is employed to reveal the multi-dimensional clustering distribution characteristics of floor instability and rupture. Through field monitoring and analysis, the evolutionary sequence of water inrush characterized by “intense mining disturbance–high KDE–water level decline–high water inrush” is identified. Given the complexity and discreteness of the quantitative indicator data, a long short-term memory (LSTM)–multi-head attention (MHA)–Bayesian optimization (BO) model is developed. Evaluation results indicate that the model demonstrates accuracy, precision, recall, and F1 score of 0.8901, 0.8476, 0.8476, and 0.8436, respectively, outperforming traditional machine learning and deep learning models. Field applications further show that the LSTM–MHA–BO model achieves coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) values of 0.924/0.971, 0.06/0.04, and 0.04/0.04, respectively, and that it successfully provides early warnings for high water inrush events in both mines.