Scholarlink

<p>This study addresses the limited application of advanced machine learning techniques in three-dimensional (3D) slope reliability analysis by evaluating the structural safety of a railway embankment in the Mokama region of Bihar, India. A comprehensive 3D slope stability assessment is conducted via 3D extensions of Bishop’s simplified method and the ordinary column method, which explicitly incorporate pore-water pressure and horizontal seismic loading. To account for spatial variability in soil properties, 20 distinct slope cases are generated, producing a dataset of 1,000 samples through Scoops3D simulations. Three artificial intelligence paradigms, artificial neural networks, genetic programming, and relevance vector machines, are employed to model the nonlinear relationships between geotechnical parameters and the computed 3D factors of safety. Comparative performance analysis reveals that the artificial neural network delivers the highest predictive accuracy, with R<sup>2</sup> values of up to 0.999 for training and 0.996 for testing, outperforming both other methods. Reliability analysis of the predicted outputs reveals that the reliability index (<i>β</i>) consistently indicates safe embankment performance, whereas the associated probability of failure (<i>P</i><sub><i>f</i></sub>) remains acceptably low under all loading scenarios. The sensitivity results indicate that cohesion is the most influential parameter for both the OCM and the BSM, followed by hydraulic conductivity and the pore pressure ratio, whereas the unit weight has the smallest effect. The primary contribution of this study lies in the integration of dual 3D limit‒equilibrium formulations with diverse machine‒learning models to enhance the predictive capability and reliability estimation for complex railway embankments, thereby offering a robust framework for future 3D geotechnical risk assessment.</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

3D reliability assessment of railway embankment under pore-water pressure and seismic loading using supervised machine learning models

  • 期刊论文

  • 发表日期：

    2026年5月15日

查看全文
  • Brijbhan Rao,
  • Divesh Ranjan Kumar,
  • Sumit Kumar,
  • Warit Wipulanusat,
  • Avijit Burman,
  • Lal Bahadur Roy

摘要

This study addresses the limited application of advanced machine learning techniques in three-dimensional (3D) slope reliability analysis by evaluating the structural safety of a railway embankment in the Mokama region of Bihar, India. A comprehensive 3D slope stability assessment is conducted via 3D extensions of Bishop’s simplified method and the ordinary column method, which explicitly incorporate pore-water pressure and horizontal seismic loading. To account for spatial variability in soil properties, 20 distinct slope cases are generated, producing a dataset of 1,000 samples through Scoops3D simulations. Three artificial intelligence paradigms, artificial neural networks, genetic programming, and relevance vector machines, are employed to model the nonlinear relationships between geotechnical parameters and the computed 3D factors of safety. Comparative performance analysis reveals that the artificial neural network delivers the highest predictive accuracy, with R2 values of up to 0.999 for training and 0.996 for testing, outperforming both other methods. Reliability analysis of the predicted outputs reveals that the reliability index (β) consistently indicates safe embankment performance, whereas the associated probability of failure (Pf) remains acceptably low under all loading scenarios. The sensitivity results indicate that cohesion is the most influential parameter for both the OCM and the BSM, followed by hydraulic conductivity and the pore pressure ratio, whereas the unit weight has the smallest effect. The primary contribution of this study lies in the integration of dual 3D limit‒equilibrium formulations with diverse machine‒learning models to enhance the predictive capability and reliability estimation for complex railway embankments, thereby offering a robust framework for future 3D geotechnical risk assessment.