Scholarlink

<p>Slope stability analysis is important for assessing landslide hazards. For this, deterministic analysis is commonly practiced. In such cases, uncertainties are disregarded and conservative point estimates are provided. To address this, probabilistic analysis is performed. Despite its advantages, such analysis is data-intensive and requires substantial computational resources for determining the Probability of Failure (<i>P</i><sub><i>f</i></sub>). However, limited research has explored computationally efficient probabilistic frameworks. Therefore, this study aims to present an efficient approach for such applications. To do this, Multivariate Adaptive Regression Splines (MARS) based surrogate modeling and Fourth Moment Normal Transformation (FMNT) based approaches were applied and compared. While surrogate modeling is well established, the efficacy of the FMNT approach has yet to be explored. With this consideration, a landslide affected rock slope located along National Highway-10 in the Darjeeling Himalaya, India, was investigated. Here, the Finite Element Method (FEM) was applied, considering Uniaxial Compressive Strength and Geological Strength Index as random variables with unknown probability density functions. Based on this, the surrogate model resulted in <i>P</i><sub><i>f</i></sub> values of 5.35% and 36.28%, while the FMNT approach resulted in <i>P</i><sub><i>f</i></sub> values of 4.55% and 23.41% for dry and wet slope mass conditions, respectively. Although the results from both approaches were comparable, training the surrogate model required 580 deterministic FEM simulations, whereas the FMNT approach required only 22 FEM simulations. This comparative evaluation reveals that the proposed FMNT approach is about 26 times more computationally efficient than the traditional surrogate modeling.</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

Probabilistic Slope Stability Assessment Using Information-Theoretic Model Selection and FEM

  • 期刊论文

  • 发表日期：

    2025年8月21日

查看全文
  • Suvam Das,
  • Ajeet Kumar Verma,
  • Anindya Pain,
  • Debi Prasanna Kanungo,
  • Koushik Pandit,
  • Shantanu Sarkar

摘要

Slope stability analysis is important for assessing landslide hazards. For this, deterministic analysis is commonly practiced. In such cases, uncertainties are disregarded and conservative point estimates are provided. To address this, probabilistic analysis is performed. Despite its advantages, such analysis is data-intensive and requires substantial computational resources for determining the Probability of Failure (Pf). However, limited research has explored computationally efficient probabilistic frameworks. Therefore, this study aims to present an efficient approach for such applications. To do this, Multivariate Adaptive Regression Splines (MARS) based surrogate modeling and Fourth Moment Normal Transformation (FMNT) based approaches were applied and compared. While surrogate modeling is well established, the efficacy of the FMNT approach has yet to be explored. With this consideration, a landslide affected rock slope located along National Highway-10 in the Darjeeling Himalaya, India, was investigated. Here, the Finite Element Method (FEM) was applied, considering Uniaxial Compressive Strength and Geological Strength Index as random variables with unknown probability density functions. Based on this, the surrogate model resulted in Pf values of 5.35% and 36.28%, while the FMNT approach resulted in Pf values of 4.55% and 23.41% for dry and wet slope mass conditions, respectively. Although the results from both approaches were comparable, training the surrogate model required 580 deterministic FEM simulations, whereas the FMNT approach required only 22 FEM simulations. This comparative evaluation reveals that the proposed FMNT approach is about 26 times more computationally efficient than the traditional surrogate modeling.