Scholarlink

<p>During high speed railway construction, shield-tunnel undercrossing frequently induces subgrade settlement, which threatens project safety and progress. Existing settlement monitoring methods struggle to provide timely early warnings due to unclear data features and inadequate long-term dependency modeling.To address this, we propose a settlement early warning method for high-speed railway subgrades based on TD Transformer. Firstly, we utilize temporal-spatial enhanced attention (TSEA) for feature extraction from high-speed railway settlement data, effectively resolving the problem of vague features post-extraction. Secondly, dynamic global temporal attention (DGTA) is employed to dynamically capture and represent the long-term dependencies of settlement data. Experimental results demonstrate that TD Transformer achieves Accuracy, Precision, Recall, and F1-Score of 93.39%, 93.10%, 93.40%, and 93.24%, respectively, outperforming other advanced settlement early warning methods for high-speed railway subgrade with relative improvements of 1.24%, 1.3%, 1.3%, and 1.27%.This method effectively forecasts subgrade settlement and exhibits significant superiority in the task of multi-factor settlement early warning for high-speed railway subgrades.</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

Settlement early warning method for high speed railway subgrades based on TD Transformer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月5日

查看全文
  • Wen Kebing,
  • Liang Qinghuai

摘要

During high speed railway construction, shield-tunnel undercrossing frequently induces subgrade settlement, which threatens project safety and progress. Existing settlement monitoring methods struggle to provide timely early warnings due to unclear data features and inadequate long-term dependency modeling.To address this, we propose a settlement early warning method for high-speed railway subgrades based on TD Transformer. Firstly, we utilize temporal-spatial enhanced attention (TSEA) for feature extraction from high-speed railway settlement data, effectively resolving the problem of vague features post-extraction. Secondly, dynamic global temporal attention (DGTA) is employed to dynamically capture and represent the long-term dependencies of settlement data. Experimental results demonstrate that TD Transformer achieves Accuracy, Precision, Recall, and F1-Score of 93.39%, 93.10%, 93.40%, and 93.24%, respectively, outperforming other advanced settlement early warning methods for high-speed railway subgrade with relative improvements of 1.24%, 1.3%, 1.3%, and 1.27%.This method effectively forecasts subgrade settlement and exhibits significant superiority in the task of multi-factor settlement early warning for high-speed railway subgrades.