摘要
钻爆法隧道洞门结构是隧道工程的重要组成部分,其洞口位置对洞口区域的开挖范围和洞门型式的确定具有显著影响,洞口位置受控因素多、依赖主观经验强。为了有效解决这些难题,引入人工智能技术,提出融合多种深度学习算法的隧道洞口位置智能决策方法。首先,通过梳理分析隧道洞口位置设计影响因素,建立隧道洞门设计数据库。针对不同类型的影响因素数据进行特征工程,在数据预处理的基础上构建融合长短期记忆网络(LSTM)、自注意力机制(Self-Attention)、卷积神经网络(CNN)和全连接神经网络(FC)的智能设计模型,实现对隧道洞口位置的智能决策。通过对比分析多种模型结构的预测效果,提出洞口位置最优预测模型LstmAttCnnNet,计算得到决定系数R^(2)达到0.910,均方根误差(RMSE)稳定在0.094。研发隧道洞口位置智能决策模块,通过BIM技术将智能决策得到的洞门长度参数进行三维展示,通过42座实际隧道洞口工程案例验证和典型工程应用,证明了决策方法的有效性和适用性。本文提出的隧道洞口位置智能预测模型首次实现了隧道洞门长度的智能决策,有效促进了钻爆法隧道智能设计技术创新,赋能钻爆法隧道智能建造。
The tunnel portal structure in drill-blasting method is a significant part of tunnel engineering.The location of the portal has a significant impact on the excavation range of the portal area and the determination of the portal type.The portal location is influenced by many factors and heavily relies on subjective experience.To effectively address these challenges,artificial intelligence technology was introduced,and an intelligent decision-making method for tunnel portal location that integrated multiple deep learning algorithms was proposed.First,by reviewing and analyzing the factors affecting tunnel portal location design,a tunnel portal design database was established.Feature engineering was applied to different types of influencing factors.Based on data preprocessing,an intelligent design model that integrated Long Short-Term Memory(LSTM)networks,Self-Attention mechanisms,Convolutional Neural Networks(CNN),and Fully Connected Neural Networks(FC)was developed to realize intelligent decision-making for tunnel portal location.By comparing and analyzing the prediction performance of various model structures,the optimal model for portal location prediction,LstmAttCnnNet,was proposed.The coefficient of determination(R^(2))was calculated to be 0.910,and the root mean square error(RMSE)remained stable at 0.094.An intelligent decision-making module for tunnel portal location was developed,and the portal length parameters obtained through intelligent decision-making were visually displayed in threedimensional format using BIM technology.The effectiveness and applicability of the decision-making method were validated through 42 actual tunnel portal engineering cases and typical engineering applications.The intelligent prediction model for tunnel portal location proposed in this paper is the first to achieve intelligent decision-making for tunnel portal length,effectively driving the innovation of intelligent design technology for tunnels using drill-blasting method and enabling the intelligent construction of these tunnels.
作者
吴佳明
戴林发宝
肖明清
杨剑
孙文昊
王峥峥
陈韶平
WU Jiaming;DAI Linfabao;XIAO Mingqing;YANG Jian;SUN Wenhao;WANG Zhengzheng;CHEN Shaoping(China Railway Siyuan Survey and Design Group Co.,Ltd.,Wuhan 430063,China;National and Local Joint Engineering Research Center of Underwater Tunneling Technology,Wuhan 430063,China;School of Infrastructure,Dalian University of Technology,Dalian 116024,China)
出处
《铁道标准设计》
北大核心
2025年第3期138-147,共10页
Railway Standard Design
基金
国家重点研发计划项目(2021YFB2600400)
中国铁建股份有限公司科技研发计划项目(2022-A02)。
关键词
隧道工程
洞口位置
深度学习
智能设计模型
智能决策
三维展示
tunnel engineering
tunnel portal location
deep learning
intelligent design model
intelligent decision-making
three-dimensional display
作者简介
吴佳明(1993-),男,工程师,2021年毕业于中国科学院大学岩土工程专业,工学博士,主要从事隧道及地下工程数智化设计与研究工作,E-mail:1345233582@qq.com。
