铁路隧道衬砌病害轻量化智能检测技术研究  

Research on Lightweight Intelligent Detection Technology of Railway Tunnel Lining Disease

作　　者：田阳 刘桂卫 崔庆国 张晨 何明峰 TIAN Yang;LIU Guiwei;CUI Qingguo;ZHANG Chen;HE Mingfeng(China Railway Design Corporation,Tianjin 300142,China)

机构地区：[1]中国铁路设计集团有限公司,天津300142

出　　处：《铁道工程学报》2025年第2期85-90,共6页Journal of Railway Engineering Society

基　　金：中国铁路设计集团有限公司科研项目(2024B0103000001)。

摘　　要：研究目的:铁路隧道衬砌质量检测工程对于隧道和铁路运营安全具有重要意义。目前,铁路隧道衬砌病害检测主要依赖人工对隧道衬砌雷达图像的病害特征进行判识,该方法判识效率低下且受主观因素影响较大。因此,为提高隧道衬砌病害检测效率,研究基于卷积神经网络的衬砌病害图像智能检测技术。进一步地,为提高该技术的工程应用能力,研究更加轻量化的隧道衬砌病害智能检测技术。研究结论:(1)通过研究隧道衬砌病害雷达图像特征,发现背景干扰较大,对病害先验知识要求较高,导致判识效率低下;(2)提出了一种衬砌欠厚判定方法,该方法可滤除大部分背景干扰且有效标记欠厚病害;(3)构建了一种新的轻量化衬砌病害检测网络模型,对比试验表明该模型可用于标记脱空、缺筋病害且具有较高的检测精度及较好的轻量化性能;(4)实际工程应用试验表明该技术在铁路隧道质量检测领域具有一定的工程实用性,可有效代替人工判识方法开展隧道衬砌病害检测工作。Research purposes:The quality inspection of railway tunnel linings is crucial for tunnel integrity and railway operational safety.Currently,the detection of tunnel lining defects primarily relies on the manual interpretation of radar images,and the process is inefficient and subject to significant subjective bias.To enhance the efficiency of defect detection,this study proposes an intelligent detection technology based on convolutional neural networks for tunnel lining defect images.Furthermore,to improve the applicability of this technology in practical engineering,we focus on developing a more lightweight tunnel lining defect detection technique.Research conclusions:(1)Analysis of radar images of tunnel lining defects revealed significant background interference and a high dependence on prior knowledge of defects,resulting in low identification efficiency.(2)A novel method for determining insufficient lining thickness was developed.This method can filter most background noise and effectively mark thickness deficiencies.(3)A new lightweight network model for detecting tunnel lining defects was constructed.Comparative experiments demonstrated that the model accurately detects voids and rebar deficiencies,achieving high detection precision with excellent lightweight performance.(4)Field trials demonstrated the practical engineering applicability of this technology in railway tunnel quality inspections,providing an effective alternative to manual identification methods for detecting tunnel lining defects.

关 键 词：铁路工程 隧道衬砌病害 雷达图像 智能检测 轻量化 

分 类 号：U212.3[交通运输工程—道路与铁道工程]