基于PSO-RBF神经网络模型的铁路隧道突水危险性评价  

Evaluation of Water Inrush Danger in Railway Tunnel Based on PSO-RBF Neural Network Model

作　　者：张鑫 靳春玲[1] 贡力[1] 魏晓悦 杜秀萍[2] ZHANG Xin;JIN Chunling;GONG Li;WEI Xiaoyue;DU Xiuping(School of Civil Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;Baoding Water Resources and Hydropower Survey and Design Institute,Baoding 071000,China)

机构地区：[1]兰州交通大学土木工程学院,兰州730070 [2]保定市水利水电勘测设计院,河北保定071000

出　　处：《铁道标准设计》2022年第10期143-148,共6页Railway Standard Design

基　　金：国家自然科学基金项目(51969011);甘肃省科技计划资助项目(20JR10RA274,20JR2RA002);甘肃省教育厅优秀研究生“创新之星”项目(2021CXZX-639)。

摘　　要：突水是铁路隧道施工过程中发生频率最高的灾害事故,为有效预防突水事故,降低隧道施工风险,保障施工人员安全。在已有研究基础上选取10个核心指标作为影响突水事故发生的判断依据,收集50组典型隧道突水实例数据作为突水危险性评价的研究样本,运用粒子群优化算法(PSO)优化径向基神经网络(RBF)后,对样本数据进行训练测试,建立PSO-RBF神经网络铁路隧道突水危险性评价模型。最后,将该模型应用于井家山隧道验证其实用性。实例研究表明:PSO-RBF模型能够准确对井家山隧道突水危险性作出判定,且与梯度下降法改进的RBF神经网络相比,PSO-RBF神经网络模型具有更高的准确率和更快的迭代速度。Water inrush is the most frequent disaster in the process of railway tunnel construction.In order to effectively prevent water inrush accidents,reduce risk in tunnel construction and ensure the safety of construction personnel,this paper,on the basis of existing researches,selects 10 core indexes as the judgment basis for the occurrence of water inrush accident,collects 50 groups of typical tunnel water inrush cases data as the research samples of water inrush risk assessment,uses particle swarm optimization algorithm(PSO)to optimize radial basis function neural network(RBF),trains and tests the sample data,and establishes the PSO-RBF neural network railway tunnel water inrush risk assessment model.Finally,the model is applied to Jingjiashan Tunnel to verify its practicability.The case study shows that the PSO-RBF model can accurately determine the risk of water inrush in Jingjiashan Tunnel,and the PSO-RBF neural network model has higher accuracy and faster iterative speed in comparison with the RBF neural network improved by the gradient descent method,.

关 键 词：铁路隧道 突水 危险性评价 粒子群算法 径向基神经网络 

分 类 号：U45[建筑科学—桥梁与隧道工程] U456.3[交通运输工程—道路与铁道工程]