Scholarlink

<p>This study is aimed at exploring the instability failure mechanisms and precursory early-warning methods of the solidified body-coal (S-C) combination bearing structure. To achieve this aim, the dynamic response characteristics of multiple acoustic emission (AE) parameters from S-C combination specimens during loading were systematically analyzed through a combination of uniaxial compression experiments and AE monitoring. Furthermore, a crack propagation stage identification model was constructed based on machine learning. The following beneficial findings were yielded: The temporal evolution patterns of energy-frequency distribution indicators (<i>Activity</i> and <i>b</i>-value) and spectral characteristics (average frequency (<i>AF</i>) and risetime/amplitude (<i>RA</i>)), derived from AE time-frequency parameters, are highly correlated with the staged crack propagation behavior, and can be regarded as key precursors of instability in rock-coal combinations. Among the machine learning-based models constructed for identifying the staged fracture propagation states, light gradient boosting machine (LGB) performs the best overall, achieving a prediction accuracy of over 85% across all stages. Such a result verifies the strong characterization capability of fused AE parameter analysis for dynamic fracture evolution. The comprehensive early warning indicator CWI(<i>t</i>) was proposed, and its weight coefficients were optimized by SHapley Additive exPlanations (SHAP) interpretability analysis, which markedly enhances the systematicity and robustness of instability risk assessment. These findings provide both a theoretical foundation and a data-driven method for the stability monitoring and early warning of the S-C combination bearing structure in continuous driving and gangue backfilling systems.</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

Characterization of acoustic emission parameters and identification of staged fracture propagation in solidified body-coal combination based on experimental and machine learning approaches

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月11日

查看全文
  • Yi Tan,
  • Hao Cheng,
  • Manchao He,
  • Wenbing Guo,
  • Erhu Bai,
  • Hui Li,
  • Weiyu Guo,
  • Yu Wang,
  • Shaopu Zhang,
  • Bingxing Chang,
  • Shicheng Hu

摘要

This study is aimed at exploring the instability failure mechanisms and precursory early-warning methods of the solidified body-coal (S-C) combination bearing structure. To achieve this aim, the dynamic response characteristics of multiple acoustic emission (AE) parameters from S-C combination specimens during loading were systematically analyzed through a combination of uniaxial compression experiments and AE monitoring. Furthermore, a crack propagation stage identification model was constructed based on machine learning. The following beneficial findings were yielded: The temporal evolution patterns of energy-frequency distribution indicators (Activity and b-value) and spectral characteristics (average frequency (AF) and risetime/amplitude (RA)), derived from AE time-frequency parameters, are highly correlated with the staged crack propagation behavior, and can be regarded as key precursors of instability in rock-coal combinations. Among the machine learning-based models constructed for identifying the staged fracture propagation states, light gradient boosting machine (LGB) performs the best overall, achieving a prediction accuracy of over 85% across all stages. Such a result verifies the strong characterization capability of fused AE parameter analysis for dynamic fracture evolution. The comprehensive early warning indicator CWI(t) was proposed, and its weight coefficients were optimized by SHapley Additive exPlanations (SHAP) interpretability analysis, which markedly enhances the systematicity and robustness of instability risk assessment. These findings provide both a theoretical foundation and a data-driven method for the stability monitoring and early warning of the S-C combination bearing structure in continuous driving and gangue backfilling systems.