Scholarlink

<p>Landslides generate continuous acoustic signals as a result of subsurface deformation and movement. These signals can be captured by an acoustic emission (AE) array system, which employs standardized active waveguide units to monitor internal slope deformations. However, the complex and stochastic interactions within the waveguides, coupled with multiple influencing factors, obscure the quantitative relationship between AE signals and actual deformation, making interpretation highly challenging. In this study, we integrate AE monitoring with machine learning (ML) techniques to classify landslide kinematic states and predict displacement. Two ML models were developed, achieving classification accuracy above 90% and displacement prediction with root-mean-square errors below 2&#xa0;mm. This approach provides a robust solution for cases where direct deformation measurements are impractical. The integration of AE arrays and ML models offers a cost-effective, real-time monitoring framework that enhances the understanding of landslide dynamics and strengthens early warning capabilities in complex geological settings.</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

Machine learning analysis of landslide subsurface deformation using acoustic emission data

  • 期刊论文

  • 发表日期：

    2026年2月20日

查看全文
  • Lizheng Deng,
  • Hongyong Yuan,
  • Guofeng Su,
  • Jianguo Chen,
  • Yang Chen,
  • Mingzhi Zhang

摘要

Landslides generate continuous acoustic signals as a result of subsurface deformation and movement. These signals can be captured by an acoustic emission (AE) array system, which employs standardized active waveguide units to monitor internal slope deformations. However, the complex and stochastic interactions within the waveguides, coupled with multiple influencing factors, obscure the quantitative relationship between AE signals and actual deformation, making interpretation highly challenging. In this study, we integrate AE monitoring with machine learning (ML) techniques to classify landslide kinematic states and predict displacement. Two ML models were developed, achieving classification accuracy above 90% and displacement prediction with root-mean-square errors below 2 mm. This approach provides a robust solution for cases where direct deformation measurements are impractical. The integration of AE arrays and ML models offers a cost-effective, real-time monitoring framework that enhances the understanding of landslide dynamics and strengthens early warning capabilities in complex geological settings.