Scholarlink

<p>Continuous monitoring of structural deformation is essential for ensuring the safety and stability of urban buildings. While traditional monitoring methods are expensive and limited in coverage, Interferometric Synthetic Aperture Radar (InSAR) provides a cost-effective and scalable alternative for structural health monitoring. However, InSAR alone is constrained by its revisit cycle and unidirectional measurement capabilities, making it insufficient for comprehensive structural health monitoring. Building upon our previously proposed InSAR-based three-dimensional (3D) displacement inversion method enhanced by domain knowledge, this study further develops a novel framework that integrates InSAR with an attention-enhanced Residual Long Short-Term Memory (Att-ResLSTM) model. The framework leverages InSAR-derived 3D displacement data and incorporates the Att-ResLSTM model to interpolate and forecast structural displacements under environmental influences. This integration enables the completion of missing displacement data within satellite revisit intervals and supports continuous time-series prediction of structural deformation, reducing the effective monitoring frequency from the typical 12-day revisit interval to a predictive resolution of 1–2&#xa0;days, which effectively solves the limitation of InSAR temporal resolution. The proposed approach has been successfully applied to a case study involving a supertall building in Chongqing, China. Experimental results show that the Att-ResLSTM model achieves high prediction accuracy across all displacement components, with test-set MAEs of 0.21&#xa0;mm, 0.26&#xa0;mm, and 0.47&#xa0;mm and corresponding R<sup>2</sup> values of 0.89, 0.90, and 0.94 for the minor, major, and vertical (U) axes, respectively. These results confirm the model’s effectiveness in enabling continuous and dynamic structural health monitoring.</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

Forecasting the InSAR-monitored 3D displacements for urban buildings using attention-enhanced LSTM

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Ya-Nan Du,
  • Jin-Peng Feng,
  • Jia-Yi Ding,
  • De-Cheng Feng

摘要

Continuous monitoring of structural deformation is essential for ensuring the safety and stability of urban buildings. While traditional monitoring methods are expensive and limited in coverage, Interferometric Synthetic Aperture Radar (InSAR) provides a cost-effective and scalable alternative for structural health monitoring. However, InSAR alone is constrained by its revisit cycle and unidirectional measurement capabilities, making it insufficient for comprehensive structural health monitoring. Building upon our previously proposed InSAR-based three-dimensional (3D) displacement inversion method enhanced by domain knowledge, this study further develops a novel framework that integrates InSAR with an attention-enhanced Residual Long Short-Term Memory (Att-ResLSTM) model. The framework leverages InSAR-derived 3D displacement data and incorporates the Att-ResLSTM model to interpolate and forecast structural displacements under environmental influences. This integration enables the completion of missing displacement data within satellite revisit intervals and supports continuous time-series prediction of structural deformation, reducing the effective monitoring frequency from the typical 12-day revisit interval to a predictive resolution of 1–2 days, which effectively solves the limitation of InSAR temporal resolution. The proposed approach has been successfully applied to a case study involving a supertall building in Chongqing, China. Experimental results show that the Att-ResLSTM model achieves high prediction accuracy across all displacement components, with test-set MAEs of 0.21 mm, 0.26 mm, and 0.47 mm and corresponding R2 values of 0.89, 0.90, and 0.94 for the minor, major, and vertical (U) axes, respectively. These results confirm the model’s effectiveness in enabling continuous and dynamic structural health monitoring.