Scholarlink

<p>The aim of this study is to examine circular tunnel stability by investigating the influence of the spatial variability of rock masses by using the Hoek–Brown model, random field theory, and random adaptive finite element limit analysis (RAFELA). The analysis investigates the influences of six input parameters, including the cover depth ratio (<i>C/D</i>), the geological strength index dependent on rock quality (<i>GSI</i>), the yield parameter (<i>m</i><sub><i>i</i></sub>), the mean of variation (<i>COVσ</i><sub><i>ci</i></sub>), the dimensionless correlation length (<i>Θ</i>), and the specified factor of safety (<i>FoS</i>), on the probabilistic analysis of circular tunnel stability. The mean stability number <i>μN</i><sub><i>ran</i></sub> increases, and the <i>PoF</i> decreases with increasing correlation length. The failure patterns of circular tunnels in rock under deterministic analysis and stochastic analysis have been examined. Furthermore, the eXtreme Gradient Boosting (XGB) method was selected to consider the relationships between the investigated parameters and the stability factor. A hybrid machine-learning framework is proposed that integrates an optimization algorithm, namely, particle swarm optimization (PSO), into the XGB algorithm to increase its performance. The efficiency of the PSO-XGB model is suggested as the best hybrid XGB model (R<sup>2</sup> = 99.61%) for predicting the failure probability of a circular tunnel.</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

Application of particle swarm optimization‐XGBoost for tunnel stability prediction by considering the random field of Hoek–Brown strength parameter

  • 期刊论文

  • 发表日期：

    2025年3月13日

查看全文
  • Thanachon Promwichai,
  • Duy Tan Tran,
  • Suraparb Keawsawasvong,
  • Pitthaya Jamsawang

摘要

The aim of this study is to examine circular tunnel stability by investigating the influence of the spatial variability of rock masses by using the Hoek–Brown model, random field theory, and random adaptive finite element limit analysis (RAFELA). The analysis investigates the influences of six input parameters, including the cover depth ratio (C/D), the geological strength index dependent on rock quality (GSI), the yield parameter (mi), the mean of variation (COVσci), the dimensionless correlation length (Θ), and the specified factor of safety (FoS), on the probabilistic analysis of circular tunnel stability. The mean stability number μNran increases, and the PoF decreases with increasing correlation length. The failure patterns of circular tunnels in rock under deterministic analysis and stochastic analysis have been examined. Furthermore, the eXtreme Gradient Boosting (XGB) method was selected to consider the relationships between the investigated parameters and the stability factor. A hybrid machine-learning framework is proposed that integrates an optimization algorithm, namely, particle swarm optimization (PSO), into the XGB algorithm to increase its performance. The efficiency of the PSO-XGB model is suggested as the best hybrid XGB model (R2 = 99.61%) for predicting the failure probability of a circular tunnel.