铁道标准设计2025，Vol.69Issue(3)：138-147,10.DOI:10.13238/j.issn.1004-2954.202407170003
基于深度学习的隧道洞口位置智能决策方法
Intelligent Decision-Making Method for Tunnel Portal Location Based on Deep Learning
摘要
Abstract
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(R2)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 three-dimensional 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.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Key words
tunnel engineering/tunnel portal location/deep learning/intelligent design model/intelligent decision-making/three-dimensional display分类
交通工程引用本文复制引用
吴佳明,戴林发宝,肖明清,杨剑,孙文昊,王峥峥,陈韶平..基于深度学习的隧道洞口位置智能决策方法[J].铁道标准设计,2025,69(3):138-147,10.基金项目
国家重点研发计划项目(2021YFB2600400) （2021YFB2600400）
中国铁建股份有限公司科技研发计划项目(2022-A02) （2022-A02）
