Scholarlink

<p>This study presents a novel machine learning framework for predicting the probability of failure in circular tunnels constructed within spatially random anisotropic clay under plane strain conditions. This research leverages random field theory (RFT) with Monte Carlo simulations to establish a random adaptive finite element limit analysis (RAFELA) approach, enabling the development of limit state solutions for tunnels. Statistical analysis is then employed to assess the factor of safety and determine the probability of failure. This work integrates RAFELA results with machine learning results by training four tree-based algorithms (Random Forest, XGBoost, AdaBoost, and CatBoost) on a comprehensive dataset of 1536 numerical solutions. These ML models incorporate five key input parameters including (1) cover depth ratio, (2) anisotropic strength ratio, (3) coefficient of variation, (4) dimensional spatial correlation length, and (5) factor of safety. The performance of each algorithm is rigorously evaluated using various accuracy metrics. The resulting ML models function as powerful surrogate models that can accurately predict the probability of failure for circular tunnels constructed in spatially random anisotropic clay.</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

Undrained Stability Prediction of Circular Tunnels in Spatially Random Anisotropic Clays Using Tree-Based Machine Learning Models

  • 期刊论文

  • 发表日期：

    2025年10月4日

查看全文
  • Kongtawan Sangjinda,
  • Natakorn Phuksuksakul,
  • Thanh Son Nguyen,
  • Suraparb Keawsawasvong,
  • Pitthaya Jamsawang

摘要

This study presents a novel machine learning framework for predicting the probability of failure in circular tunnels constructed within spatially random anisotropic clay under plane strain conditions. This research leverages random field theory (RFT) with Monte Carlo simulations to establish a random adaptive finite element limit analysis (RAFELA) approach, enabling the development of limit state solutions for tunnels. Statistical analysis is then employed to assess the factor of safety and determine the probability of failure. This work integrates RAFELA results with machine learning results by training four tree-based algorithms (Random Forest, XGBoost, AdaBoost, and CatBoost) on a comprehensive dataset of 1536 numerical solutions. These ML models incorporate five key input parameters including (1) cover depth ratio, (2) anisotropic strength ratio, (3) coefficient of variation, (4) dimensional spatial correlation length, and (5) factor of safety. The performance of each algorithm is rigorously evaluated using various accuracy metrics. The resulting ML models function as powerful surrogate models that can accurately predict the probability of failure for circular tunnels constructed in spatially random anisotropic clay.