管线渗漏破坏下地铁隧道施工坍塌风险预测  

Prediction of the collapse accident probability of urban subway tunnel construction under the condition of damaged and leaking pipelines

作　　者：王 刘保国[1] 亓轶 WANG Yan;LIU Baoguo;QI Yi(School of Civil Engineering, Beijing Jiaotong University, Beijing 100044, China)

机构地区：[1]北京交通大学土木建筑工程学院,北京100044

出　　处：《岩石力学与工程学报》2018年第A01期3432-3440,共9页Chinese Journal of Rock Mechanics and Engineering

基　　金：中央高校基本科研业务费专项资金资助项目(2016YJS114);国家自然科学基金资助项目(71771020)~~

摘　　要：针对我国地铁建设坍塌事故多发的现实情况,提出基于数值模拟–人工神经网络–蒙特卡罗原理耦合的坍塌事故风险预测方法,在管线渗漏破坏条件下对坍塌事故风险进行预测;在事故统计调研的基础上,得到管线渗漏破坏是坍塌事故发生的主要原因;通过有限差分流固耦合数值模拟,计算得到管线分别位于隧道正上方、隧道正上方偏右5 m和偏右10 m三种管线位置地表最大沉降值,当管线未渗漏破坏时,最大沉降均位于隧道正上方地表,最大沉降值分别为0.012 85,0.016 05和0.018 53 m,当管线渗漏破坏时,最大沉降分别位于隧道正上方地表和隧道上方偏右约2 m处地表,最大沉降值分别为0.028 75,0.027 17和0.021 8 m;将数值模拟结果作为神经网络的训练样本和检验样本,采用RBF神经网络建立基本参数与地表沉降值的非线性映射关系用以代替蒙特卡罗原理功能函数,采用地表沉降值作为反映坍塌风险倾向的数值参考,根据蒙特卡罗原理,计算得到管线渗漏破坏时3种位置的隧道开挖坍塌风险,分别为36.75%,25.08%和接近0;论文研究内容及成果可以为类似地铁隧道建设风险控制提供借鉴。Based on the reality of frequent collapse accidents in subway construction in China,a method of collapse risk prediction based on numerical simulation,artificial neural networks and Monte Carlo method is put forward. The risk of collapse accident is predicted under the condition of pipeline leakage and damage. On the basis of investigation of accident statistics,the main reason that resulted in collapse accidents is leakage of pipelines. Through the method of finite difference fluid-solid coupling numerical simulation,the maximum ground settlement values of the three different pipeline locations,namely,just above the tunnel,above the tunnel 5 m on the right and above the tunnel 10 m on the right,are calculated. When pipelines are not damaged and leaking,the maximum ground settlement is located at the surface just above the tunnel and the values are 0.012 85,0.016 05 and 0.018 53 m,respectively. When pipelines are damaged and leaking,the maximum ground settlement located at the surface just above the tunnel and above the tunnel about 2 m on the right. The maximum ground settlement values are 0.028 75,0.027 17 and 0.021 8 m,respectively. The numerical simulation results are used as the training and test samples of neural network,and the non-linear mapping relationship between the basic parameters and the ground settlement is established by RBF neural network,which is used to replace the performance function of Monte-Carlo method. According to Monte-Carlo method,the probability of collapse risk of the three locations under the condition of damaged and leaking pipelines is calculated when the tunnel is excavated. The probabilities are 36.75%,25.08% and close to 0. This research can provide reference for similar risk control of subway tunnel construction.

关 键 词：隧道工程 事故调查 数值模拟 RBF神经网络 蒙特卡罗 坍塌风险 

分 类 号：U45[建筑科学—桥梁与隧道工程]