基于BP神经网络的铁路长深隧道地应力三维反演技术研究与应用  

Research & Application of 3D Inversion of In-situ Stress for Long and Deep Railway Tunnel based on BP Neural Network

作　　者：张晨 王恒松 刘桂卫 赵建峰 Zhang Chen;Wang Hengsong;Liu Guiwei;Zhao Jianfeng(China Railway Design Corporation,Tianjin 300251,China)

机构地区：[1]中国铁路设计集团有限公司,天津300251

出　　处：《市政技术》2022年第10期143-149,共7页Journal of Municipal Technology

基　　金：中国铁路设计集团有限公司重点科研课题资助项目(No.2021A240407)。

摘　　要：针对区域地应力评估难度大,难以通过实测方法获得区域地应力真实情况的问题,在实测点状地应力数据基础上,使用优化的BP神经网络与数值模拟相结合的方法,对浩吉铁路阳山隧道工程范围内岩体的区域地应力状态进行了预测与模拟,实现了对区域地应力水平更贴近现实情况的预测和评估。通过对比优化的BP神经网络与回归分析2种方法可知,前者能够更好地克服地应力反演结果离散性大的问题。通过实例验证可知,采用优化的BP神经网络与数值模拟相结合的方法进行区域地应力评价是可行的。In view of the difficulty of assessment and obtaining the real situation of regional in-situ stress, based on the measured point in-situ stress data, the optimized BP neural network and numerical simulation method are combined to predict and simulate the regional in-situ stress state of rock mass within the scope of Yangshan Tunnel Project of Haolebaoji-Ji’an Railway. The regional in-situ stress level is predicted and evaluated more closer to the reality. By comparing the optimized BP neural network with regression analysis, it can be seen that the former can better overcome the problem of large discreteness of in-situ stress inversion results. It is feasible to use the optimized BP neural network and numerical simulation method to evaluate regional in-situ stress.

关 键 词：铁路隧道 地应力 BP神经网络 数值模拟 回归分析 三维反演 

分 类 号：U459.1[建筑科学—桥梁与隧道工程]