Scholarlink

<p>Due to the influences of rainfall and reservoir water level fluctuations, the stability of slopes along the banks of the Three Gorges Reservoir area changes with the long-term dynamics of external disaster-causing factors. Efficiently and accurately assessing the time-dependent reliability analysis of reservoir landslides remains a challenging task. From the perspective of time series prediction, this study proposes a method based on deep learning models integrated with the attention mechanism to optimize the time-dependent reliability analysis of reservoir bank landslides. This study systematically examines the predictive performance before and after the integration of the attention mechanism with three deep learning algorithms, namely Convolutional Neural Network (CNN), Long Short-Term Memory network (LSTM), and Gate Recurrent Unit (GRU). It explores the predictive efficacy of the Neural basis expansion analysis for interpretable time series forecasting (N-BEATS) model, a structure distinct from mainstream deep learning models, in the analysis of landslide time-dependent failure probability. This study also analyzes the robustness and predictive randomness of different deep learning models. The results show that the developed attention-based deep learning models enhance the time-dependent reliability analysis of reservoir slopes. The Attention-based GRU model achieved the best predictive performance, and the Attention-based LSTM model gained the most significant improvement in predictive ability. The N-BEATS model exhibits broad application prospects in the field of time series prediction of landslide disasters.</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

Advancing dynamic reliability assessment of reservoir slopes using attention-based neural networks

  • 期刊论文

  • 发表日期：

    2025年5月29日

查看全文
  • Lin Wang,
  • Kangjie Yang,
  • Yang Zhou,
  • Junzhi Liu,
  • Haoran Hu

摘要

Due to the influences of rainfall and reservoir water level fluctuations, the stability of slopes along the banks of the Three Gorges Reservoir area changes with the long-term dynamics of external disaster-causing factors. Efficiently and accurately assessing the time-dependent reliability analysis of reservoir landslides remains a challenging task. From the perspective of time series prediction, this study proposes a method based on deep learning models integrated with the attention mechanism to optimize the time-dependent reliability analysis of reservoir bank landslides. This study systematically examines the predictive performance before and after the integration of the attention mechanism with three deep learning algorithms, namely Convolutional Neural Network (CNN), Long Short-Term Memory network (LSTM), and Gate Recurrent Unit (GRU). It explores the predictive efficacy of the Neural basis expansion analysis for interpretable time series forecasting (N-BEATS) model, a structure distinct from mainstream deep learning models, in the analysis of landslide time-dependent failure probability. This study also analyzes the robustness and predictive randomness of different deep learning models. The results show that the developed attention-based deep learning models enhance the time-dependent reliability analysis of reservoir slopes. The Attention-based GRU model achieved the best predictive performance, and the Attention-based LSTM model gained the most significant improvement in predictive ability. The N-BEATS model exhibits broad application prospects in the field of time series prediction of landslide disasters.