摘要
为实现富水区隧道建设的参数反馈优化,基于机器学习和智能优化算法建立了流固耦合条件下的隧道参数反分析方法。首先,建立工程数值计算模型,采用流固耦合模型进行求解;以正交设计分析的形式获得不同岩体参数组合条件下的围岩位移及孔隙水压分布情况,建立机器学习所需的学习样本,其中输入组为各岩体参数集,输出组为围岩位移及孔隙水压;其次,采用高斯回归过程获取学习样本所蕴含的非线性映射关系,并通过差异进化算法优化这个过程所涉及的关键参数集;然后,根据所建立的回归模型,根据目标区域的实测位移及孔隙水压,再次采用差异进化算法对岩体参数进行寻优计算,获得参数反分析结果。最后,将反演值与实测值进行对比,验证反分析得到的围岩参数的可靠性。
In order to achieve feedback optimization of parameters under tunnel construction in rich water zones,a feedback analysis method for tunnel parameters in fluid-solid coupling conditions was established based on the machine learning and intelligent optimization algorithm.Firstly,the numerical calculation model was modelled and solved by the fluid-solid coupling model.The displacement of adjoining rock and the distribution of pore water pressure in different combination of rock parameters were obtained by orthogonal design analysis,and the learning samples for machine learning were established,in which the input group was consisted of a set of rock parameters and the output group was consisted of adjoining rock displacement and pore water pressure.Then,Gaussian Regression Process was used to obtain the non-linear mapping relation contained the learning samples,and differential evolution algorithm was used to optimize the key parameters involved in the process.Furthermore,according to the established regression model and the measured displacement and pore water pressure in the target zone,the differential evolution algorithm was used to optimize the rock parameters and the parameter feedback analysis results were obtained.Finally,the inverse values were compared with the actual measured values,and the reliability of the adjoining rock parameters obtained from the feedback analysis was verified.
作者
吴招锋
WU Zhaofeng(Nanchang Rail Transit Group Co.,Ltd.,Nanchang 330000,China)
出处
《工业建筑》
CSCD
北大核心
2021年第2期140-145,205,共7页
Industrial Construction
关键词
隧道工程
流固耦合
智能反馈分析
高斯过程
差异进化算法
富水区隧道
tunnel engineering
fluid-solid coupling
intelligent feedback analysis
Gaussian Process
difference evolution algorithm
tunnel in rich water zones
