基于NAR神经网络和R/S分析法的隧道围岩变形预测分析  

Prediction and analysis of tunnel surrounding rock deformationbased on NAR neural network and R/S analysis method

作　　者：陈杨 徐浩博 赵明慧 CHEN Yang;XU Haobo;ZHAO Minghui(Shandong Yimeng Design Consulting Co.,Ltd.,Linyi 276000,Shandong,China)

机构地区：[1]山东沂蒙设计咨询有限公司,山东临沂276000

出　　处：《工程建设》2024年第5期31-36,共6页Engineering Construction

摘　　要：为研究隧道围岩变形非线性特点,采用NAR神经网络和R/S分析法,对隧道围岩变形量和变形趋势进行分析。通过NAR神经网络对变形监测样本进行误差分析,认为NAR神经网络对围岩变形短期预测时的误差小精度高。运用R/S分析法对各变形时间序列进行重标极差分析,获得各时序的Hurst指数,分析其与围岩变形趋势的关系,并通过Hurst指数对隧道围岩变形趋势进行判定。结果表明:算例中的断面围岩变形仍然会呈增长趋势,但增长幅度在减小,且水平收敛的趋势性强于拱顶沉降,说明前者受随机扰动影响较小,后期稳定性相对更高。通过运用R/S分析法进行时间序列分析,不仅为围岩变形趋势预判提供了Hurst指数判据,同时也为围岩稳定性分析及治理提供了一种依据。In order to study the nonlinear characteristics of tunnel surrounding rock deformation,NAR neural network and R/S analysis method are used to analyze the deformation amount and deformation trend of tunnel surrounding rock.Through the error analysis of deformation monitoring samples by NAR neural network,it is considered that NAR neural network has small error and high accuracy in short-term prediction of surrounding rock deformation.The R/S analysis method is used to conduct rescaled range analysis of each deformation time series,and the Hurst index of each time series is obtained,the relationship between it and the deformation trend of the surrounding rock is analyzed,and the deformation trend of the tunnel surrounding rock is judged by the Hurst index.The results show that the deformation of the surrounding rock in the section in the example will still show an increasing trend,but the growth range is decreasing,and the trend of horizontal convergence is stronger than that of the vault settlement,indicating that the former is less affected by random disturbance,and the stability is relatively higher in the later stage.By using R/S analysis method to analyze time series,it not only provides Hurst index criterion for predicting deformation trend of surrounding rock,but also provides a basis for stability analysis and treatment of surrounding rock.

关 键 词：隧道围岩 变形预测 R/S分析法 NAR神经网络 

分 类 号：TU973.23[建筑科学—结构工程]