Scholarlink

<p>The influence of underground excavation on urban areas constitutes a critical issue in tunnel engineering. This paper strives to establish a machine learning algorithm to predict maximum tunnel induced surface displacement. The algorithm was developed using FLAC3D and validated under actual conditions in the twin tunnel of the Seville metro line. A dataset of 526 simulations of underground excavation with Earth Pressure Balance (EPB) was used to predict the maximum surface displacement using machine learning techniques. Five machine learning methods to evaluate the significance of input data variables, as soil properties, tunnel depth, face pressure or grout pressure of EPB, proving the most accurate models, were Gradient Boost and XGBoost. Additionally, the feature importance analysis conducted using Random Forest indicated that soil properties play a crucial role in the prediction process. The XGBoost model's effectiveness in predicting surface displacement has been confirmed through validation on real monitored data from Seville metro tunnel Line 1. The percentage error of the calculated values was compared with the real vertical surface movements obtained, and it varied from 3.24 to 10.66%. This study develops a practical approach to improving construction planning for future excavations in Seville, making them safer and more efficient.</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

Predicting Maximum Surface Displacement from Mechanized Twin Tunnel Excavation in Seville Using Machine Learning and FLAC3D Simulation

  • 期刊论文

  • 发表日期：

    2025年1月3日

查看全文
  • Maziyar Bahri,
  • Rocío Romero-Hernández,
  • Emilio J. Mascort-Albea,
  • Cristina Soriano-Cuesta,
  • Antonio Jaramillo-Morilla

摘要

The influence of underground excavation on urban areas constitutes a critical issue in tunnel engineering. This paper strives to establish a machine learning algorithm to predict maximum tunnel induced surface displacement. The algorithm was developed using FLAC3D and validated under actual conditions in the twin tunnel of the Seville metro line. A dataset of 526 simulations of underground excavation with Earth Pressure Balance (EPB) was used to predict the maximum surface displacement using machine learning techniques. Five machine learning methods to evaluate the significance of input data variables, as soil properties, tunnel depth, face pressure or grout pressure of EPB, proving the most accurate models, were Gradient Boost and XGBoost. Additionally, the feature importance analysis conducted using Random Forest indicated that soil properties play a crucial role in the prediction process. The XGBoost model's effectiveness in predicting surface displacement has been confirmed through validation on real monitored data from Seville metro tunnel Line 1. The percentage error of the calculated values was compared with the real vertical surface movements obtained, and it varied from 3.24 to 10.66%. This study develops a practical approach to improving construction planning for future excavations in Seville, making them safer and more efficient.