Scholarlink

<p>Addressing the issue of surface settlement (SS) resulting from twin tunnel construction is a complex problem that requires the fine-tuning of numerous factors influenced by soil characteristics, tunnel geometry, and excavation methods. Modern solutions, such as machine learning (ML) and optimization algorithms, offer promising pathways to tackle this problem. This research investigates the efficacy of utilizing the Coati optimization algorithm (COA), particle swarm optimization (PSO), and Bayesian Optimization (BO) in conjunction with tree-based ML models like random forest (RF), adaptive boost (ADABoost), gradient boosting tree (GBT), extreme gradient boost (XGBoost), light gradient boosting machine (LGBM), and categorical boosting (CATBoost) to reduce SS during twin tunnel excavation. The results indicate that COA achieves the minimum SS in fewer than 10 generations for all tree-based ML algorithms, suggesting it is the most efficient algorithm for identifying the minimum SS compared to the others. Additionally, the study reveals that three different optimization algorithms (COA, PSO, and BO) combined with GBT identified the lowest SS values of 0.94&#xa0;mm, 0.88&#xa0;mm, and 0.82&#xa0;mm, respectively, all less than 1.00&#xa0;mm. This demonstrates that GBT, when paired with optimization algorithms, can achieve the lowest SS caused by twin tunnels. This study underscores the effectiveness of optimization techniques in tunnel construction, highlighting their potential in addressing engineering challenges related to SS reduction.</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

Optimizing Twin Tunnel Excavation: Machine Learning and Algorithmic Solutions for Surface Settlement Reduction

  • 期刊论文

  • 发表日期：

    2025年1月31日

查看全文
  • Chia Yu Huat,
  • Danial Jahed Armaghani,
  • Hadi Fattahi,
  • Xuzhen He,
  • Haleh Rasekh,
  • Pijush Samui

摘要

Addressing the issue of surface settlement (SS) resulting from twin tunnel construction is a complex problem that requires the fine-tuning of numerous factors influenced by soil characteristics, tunnel geometry, and excavation methods. Modern solutions, such as machine learning (ML) and optimization algorithms, offer promising pathways to tackle this problem. This research investigates the efficacy of utilizing the Coati optimization algorithm (COA), particle swarm optimization (PSO), and Bayesian Optimization (BO) in conjunction with tree-based ML models like random forest (RF), adaptive boost (ADABoost), gradient boosting tree (GBT), extreme gradient boost (XGBoost), light gradient boosting machine (LGBM), and categorical boosting (CATBoost) to reduce SS during twin tunnel excavation. The results indicate that COA achieves the minimum SS in fewer than 10 generations for all tree-based ML algorithms, suggesting it is the most efficient algorithm for identifying the minimum SS compared to the others. Additionally, the study reveals that three different optimization algorithms (COA, PSO, and BO) combined with GBT identified the lowest SS values of 0.94 mm, 0.88 mm, and 0.82 mm, respectively, all less than 1.00 mm. This demonstrates that GBT, when paired with optimization algorithms, can achieve the lowest SS caused by twin tunnels. This study underscores the effectiveness of optimization techniques in tunnel construction, highlighting their potential in addressing engineering challenges related to SS reduction.