Scholarlink

<p>Responding to the upsurge in demand for freight transportation, exclusively heavy-haul rail embankments (HHRE) were constructed to carry trains with high gross weights, huge axle loads, or unusually high traffic volumes. However, to facilitate these demands smoothly, the embankment needs a proper investigation. This study reports the slope stability assessment of the 12.293 m high HHRE using AI-based three distinctive ML models, <i>viz</i>., recurrent neural network (RNN), long short-term memory (LSTM), and bi-directional LSTM (Bi-LSTM). For this purpose, the SLOPE/W module was used as a deterministic analysis, and the FOS was calculated using Bishop’s simplified method; furthermore, three distinct ML models were used for the prediction of the FOS for the proposed embankment. Following the construction of the model, multiple key performance indicators were used to map the model. The results show that the Bi-LSTM outperformed the other developed model in terms of <i>TIC</i> = 0.0016 and <i>R</i><sup>2</sup> = 0.9995 during the development stage, while <i>TIC</i> = 0.0016 and <i>R</i><sup>2</sup> = 0.9996 during the validation stage. Additionally, the AIC value has also been calculated to find the relative superiority of all the developed models. This result suggests that Bi-LSTM achieves the lowest AIC value (i.e., –14657.78 in training and –9818.31 for testing) compared to other models, which indicates an outstanding fit and is universal. Furthermore, sensitivity analysis was investigated using the cosine amplitude method for the embankment fill input parameters to the output. The results indicate that unit weight has the highest influence on the FOS of the proposed embankment.</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

Effect of pore water pressure on slope stability in a heavy-haul railway embankment using a deep learning approach

  • 期刊论文

  • 发表日期：

    2025年4月25日

查看全文
  • Md Shayan Sabri,
  • Furquan Ahmad,
  • Pijush Samui,
  • Amit Kumar Verma,
  • Pradeep U Kurup

摘要

Responding to the upsurge in demand for freight transportation, exclusively heavy-haul rail embankments (HHRE) were constructed to carry trains with high gross weights, huge axle loads, or unusually high traffic volumes. However, to facilitate these demands smoothly, the embankment needs a proper investigation. This study reports the slope stability assessment of the 12.293 m high HHRE using AI-based three distinctive ML models, viz., recurrent neural network (RNN), long short-term memory (LSTM), and bi-directional LSTM (Bi-LSTM). For this purpose, the SLOPE/W module was used as a deterministic analysis, and the FOS was calculated using Bishop’s simplified method; furthermore, three distinct ML models were used for the prediction of the FOS for the proposed embankment. Following the construction of the model, multiple key performance indicators were used to map the model. The results show that the Bi-LSTM outperformed the other developed model in terms of TIC = 0.0016 and R2 = 0.9995 during the development stage, while TIC = 0.0016 and R2 = 0.9996 during the validation stage. Additionally, the AIC value has also been calculated to find the relative superiority of all the developed models. This result suggests that Bi-LSTM achieves the lowest AIC value (i.e., –14657.78 in training and –9818.31 for testing) compared to other models, which indicates an outstanding fit and is universal. Furthermore, sensitivity analysis was investigated using the cosine amplitude method for the embankment fill input parameters to the output. The results indicate that unit weight has the highest influence on the FOS of the proposed embankment.