Scholarlink

<p>Urban subway construction affects the stability of the ground, and land subsidence has become the most important disaster derived from subway construction. It is of great significance to timely and effectively monitor the ground subsidence along the subway, analyse the distribution characteristics of subsidence, and build a&#xa0;deformation prediction model for preventing secondary disasters. Based on PS-InSAR technology, the 31-phase Sentinel-1A images of different frames of the same track, along with GACOS products, are used to obtain the deformation rate and accumulated settlement information of three subway lines in Shijiazhuang City from March&#xa0;14, 2017 to September&#xa0;24, 2019, and the deformation gradient distribution characteristics of the lines are analysed. Considering the interaction of the spatial neighbourhood monitoring points, the CNN-LSTM-Attention model is established to predict the multi-feature settlement time series. The experimental results demonstrate that the proposed model achieved RMSE, MAPE, and MAE values of 1.758 mm, 5.51%, and 1.405 mm, respectively, outperforming other prediction models. PS-InSAR combined with network model can effectively realise the identification, extraction and prediction of deformation information along the subway line, which can provide data and technical support for the monitoring and safety plan formulation along the subway line.</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

Combined PS-InSAR and Deep Learning CNN-LSTM-Attention Model for Deformation Prediction Along Subway Lines: a Case Study in Shijiazhuang, China

  • 期刊论文

  • 发表日期：

    2026年4月23日

查看全文
  • Xinpeng Diao,
  • Zhenmin Hu,
  • Wanyu Xia,
  • Chunfeng Han,
  • Yuelong Zhao,
  • Deyang Yu,
  • Fei Xu

摘要

Urban subway construction affects the stability of the ground, and land subsidence has become the most important disaster derived from subway construction. It is of great significance to timely and effectively monitor the ground subsidence along the subway, analyse the distribution characteristics of subsidence, and build a deformation prediction model for preventing secondary disasters. Based on PS-InSAR technology, the 31-phase Sentinel-1A images of different frames of the same track, along with GACOS products, are used to obtain the deformation rate and accumulated settlement information of three subway lines in Shijiazhuang City from March 14, 2017 to September 24, 2019, and the deformation gradient distribution characteristics of the lines are analysed. Considering the interaction of the spatial neighbourhood monitoring points, the CNN-LSTM-Attention model is established to predict the multi-feature settlement time series. The experimental results demonstrate that the proposed model achieved RMSE, MAPE, and MAE values of 1.758 mm, 5.51%, and 1.405 mm, respectively, outperforming other prediction models. PS-InSAR combined with network model can effectively realise the identification, extraction and prediction of deformation information along the subway line, which can provide data and technical support for the monitoring and safety plan formulation along the subway line.