Scholarlink

<p>To improve the prediction of water inrush risk in water-rich sections of deep-buried TBM tunnels, this study developed a K-means-PCA-machine-learning framework using five causative factors, namely groundwater level, formation lithology, unfavorable geology, rock layer dip angle, and negative terrain area ratio. Principal component analysis was used for dimensionality reduction, and Random Forest (RF) and Support Vector Machine (SVM) models were established for comparison. The results show that the RF model achieved better predictive performance than the SVM model for the current case database. The proposed framework provides a practical reference for risk assessment and construction safety management in deep-buried TBM tunnels. Owing to the limited sample size and the lack of independent validation data, the present results should be interpreted as case-specific. Future work will focus on improving model interpretability, uncertainty analysis, and integration with real-time geophysical monitoring data.</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 and application of water inrush in water-rich area of TBM construction deep buried tunnel

  • 期刊论文

  • 发表日期：

    2026年4月23日

查看全文
  • Fa-yong Gao,
  • Kun Yang,
  • Yan-qiang Cheng,
  • Yong-tao Ma,
  • Kang Fu,
  • Ming-yang Tang,
  • Jin-rui Duan

摘要

To improve the prediction of water inrush risk in water-rich sections of deep-buried TBM tunnels, this study developed a K-means-PCA-machine-learning framework using five causative factors, namely groundwater level, formation lithology, unfavorable geology, rock layer dip angle, and negative terrain area ratio. Principal component analysis was used for dimensionality reduction, and Random Forest (RF) and Support Vector Machine (SVM) models were established for comparison. The results show that the RF model achieved better predictive performance than the SVM model for the current case database. The proposed framework provides a practical reference for risk assessment and construction safety management in deep-buried TBM tunnels. Owing to the limited sample size and the lack of independent validation data, the present results should be interpreted as case-specific. Future work will focus on improving model interpretability, uncertainty analysis, and integration with real-time geophysical monitoring data.