Scholarlink

<p>The design of support structures for large-span metro stations constructed by the arch cover method in upper-soft and lower-hard strata presents a complex, multi-constrained optimization problem. Conventional design approaches often struggle to balance stringent deformation control and cost-effectiveness. This study proposed a hybrid intelligent optimization framework to address this challenge. First, a comprehensive dataset of 300 numerical simulation cases was systematically generated using a Latin hypercube sampling design. Based on this dataset, an XGBoost-based surrogate model was established. Its hyperparameters were fine-tuned using the Quadruple Parameter Adaptation Growth Optimizer (QAGO), which significantly improved its predictive accuracy. Subsequently, a novel hybrid metaheuristic algorithm was proposed by adaptively combining the Slime Mould Algorithm (SMA) and the Honey Badger Algorithm (HBA), and further enhanced with a chaotic mapping strategy to bolster its global search capabilities. This optimizer utilized the surrogate model to identify cost-effective support parameters that satisfied predefined safety constraints. Performance evaluation showed that the QAGO-tuned XGBoost model’s coefficient of determination (R<sup>2</sup>) for surface settlement prediction increased from a baseline of 0.320 to 0.723. The chaos-enhanced SMA-HBA algorithm consistently outperformed standalone metaheuristic algorithms, generating support schemes that were validated through numerical simulations as both safe and economical.</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

An Optimization Model for Support Parameters of Arch Cover Method in Underground Station Excavation Based on Gradient-Boosted Regression and Metaheuristic Optimization

  • 期刊论文

  • 发表日期：

    2026年1月3日

查看全文
  • Pengtao Chen,
  • Junru Zhang,
  • Jianchi Ma,
  • Junfeng Wan,
  • Jimeng Feng,
  • Tong Pan,
  • Bo Wang,
  • Kaimeng Ma

摘要

The design of support structures for large-span metro stations constructed by the arch cover method in upper-soft and lower-hard strata presents a complex, multi-constrained optimization problem. Conventional design approaches often struggle to balance stringent deformation control and cost-effectiveness. This study proposed a hybrid intelligent optimization framework to address this challenge. First, a comprehensive dataset of 300 numerical simulation cases was systematically generated using a Latin hypercube sampling design. Based on this dataset, an XGBoost-based surrogate model was established. Its hyperparameters were fine-tuned using the Quadruple Parameter Adaptation Growth Optimizer (QAGO), which significantly improved its predictive accuracy. Subsequently, a novel hybrid metaheuristic algorithm was proposed by adaptively combining the Slime Mould Algorithm (SMA) and the Honey Badger Algorithm (HBA), and further enhanced with a chaotic mapping strategy to bolster its global search capabilities. This optimizer utilized the surrogate model to identify cost-effective support parameters that satisfied predefined safety constraints. Performance evaluation showed that the QAGO-tuned XGBoost model’s coefficient of determination (R2) for surface settlement prediction increased from a baseline of 0.320 to 0.723. The chaos-enhanced SMA-HBA algorithm consistently outperformed standalone metaheuristic algorithms, generating support schemes that were validated through numerical simulations as both safe and economical.