地铁隧道土体参数优先级表和神经网络反分析  

Back analysis of parameters of soil for subway tunnel with priority table and neural network

作　　者：王洪德[1,2] 朱贵东[1] 

机构地区：[1]大连交通大学土木与安全工程学院,辽宁大连116028 [2]大连交通大学隧道与地下结构工程技术研究中心,辽宁大连116028

出　　处：《中国安全科学学报》2014年第10期15-20,共6页China Safety Science Journal

基　　金：国家自然科学基金资助(U1261121/E0422);辽宁省科技厅公益基金资助(2014004027);中国铁路总公司科技研究开发计划重点课题(2014X012-D)

摘　　要：为获取准确的土体力学参数,保障隧道施工过程的安全可靠,提出一种优先级表和神经网络相结合的地铁隧道土体力学参数反分析方法。首先,采用优先级表的实时调度算法确定待反演土体力学参数;然后,采用正交试验设计方法,寻求反演参数取值的最优组合,正演获取训练和测试样本;其次,用混合遗传算法优化的BP神经网络,找出待反演参数与土体变形量的非线性映射关系;最后,将工程现场土体变形信息作为输入样本,对土体力学参数进行反演分析和灵敏度检查。结果表明,土体位移模拟值与实测值最大绝对误差为0.018 mm,最大相对误差为1.070%,符合安全规范及设计要求。To obtain accurate mechanical parameter values of soil mass,and to guarantee the safety and reliability of tunnel construction,a back analysis method was worked out for mechanical parameters of soil mass of subway tunnel based on combination of priority table and neural network. First,the real-time process algorithm of priority table was used to determine mechanical parameters of soil mass for inversion.Then,the orthogonal test design method was used to find optimized combination of back analysis parameters. The forward modeling method was used to acquire the sample for train and for test. Third,BP neural network optimized by hybrid genetic algorithm was used to find nonlinear mapping relationship between inversion parameters and soil mass deformation. Finally,soil mass deformation information from engineering field was taken as input samples,the mechanical parameters of soil mass were inversed,and sensitivities of parameters were checked. A comparison was made between measured displacement and displacement of forward simulation combining inversion results. Results show that greatest absolute error of soil mass displacement is 0. 018 mm,and the greatest relative error is 1. 07%,meeting safety standard and design demand.

关 键 词：土体力学参数 优先级表 神经网络 正交试验 混合遗传算法 反分析 

分 类 号：X913.4[环境科学与工程—安全科学]