Scholarlink

<p>To overcome the difficulty of accurately predicting water inrush hazards in tunnel construction, this study introduces a multi-source geophysical data fusion framework based on an MLP-Transformer model. Taking a highway tunnel project as an example, ground-penetrating radar, tunnel seismic prediction, and semi-airborne transient electromagnetic data were integrated to establish a risk evaluation system comprising six key indicators, with weights assigned using the analytic hierarchy process. The MLP was applied for feature encoding, and the Transformer with multi-head self-attention was employed for cross-modal deep fusion. Result indicates the predicted high-risk zones showed strong consistency with the actual geological conditions, confirming the accuracy and robustness of the proposed method. This framework offers a reliable technical basis for early warning of water inrush risks and for ensuring construction safety in tunnel 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

Research on Risk Evaluation of Tunnel Water Inrush Based on Multi-source Geophysical Exploration Data Fusion of MLP-Transformer Model

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • Guang Huo,
  • Huai-bing Wang,
  • Jin-gang Zhang,
  • Yi-guo Xue,
  • Bo-yin Fu,
  • Fan-meng Kong,
  • Zongwei Yan

摘要

To overcome the difficulty of accurately predicting water inrush hazards in tunnel construction, this study introduces a multi-source geophysical data fusion framework based on an MLP-Transformer model. Taking a highway tunnel project as an example, ground-penetrating radar, tunnel seismic prediction, and semi-airborne transient electromagnetic data were integrated to establish a risk evaluation system comprising six key indicators, with weights assigned using the analytic hierarchy process. The MLP was applied for feature encoding, and the Transformer with multi-head self-attention was employed for cross-modal deep fusion. Result indicates the predicted high-risk zones showed strong consistency with the actual geological conditions, confirming the accuracy and robustness of the proposed method. This framework offers a reliable technical basis for early warning of water inrush risks and for ensuring construction safety in tunnel engineering.