基于BP神经网络和R／S分析的隧道仰坡沉降变形预报预测-.pdf

施工技术 2014年6月上 80 CONSTRUCTION TECHNOLOGY 第43卷第11期 D01:10.7672/js2014110080 基于BP神经网络和R/S分析的隧道仰坡 沉降变形预报预测* 罗林，左昌群，赵连，唐霞 (中国地质大学（武汉）工程学院，湖北武汉430074) [摘要]隧道洞口处多为软弱岩或浮土，稳定性差，地表位移监测成为判断洞口稳定性的重要手段，因此仰坡沉降 变形预测显得格外重要.鉴于仰坡沉降变形具有很强的非线性特征，选取P神经网络对仰坡的沉降变形进行预 测，并验证其可行性，进而利用BP神经网络扩大沉降变形监测的样本.在此基础上，再利用R/S分析对新的监测 样本进行重标极差分析，分别得到隧道仰坡沉降-时间序列和变形速率-时间序列的Hust指数，并结合两项指数确 定了隧道仰坡沉降变形的趋势，为判断仰坡的稳定性及治理提供了有力依据. [关键词]隧道工程；仰坡；BP神经网络；R/S分析；沉降；预测 [中图分类号]U456.31 [文献标识码]A [文章编号]1002-8498(2014)11-0080-05 Settlement Deformation Prediction of the Front Slope in Tunnel Based on the BP Neural Network and R/S Analysis Luo Lin Zuo Changqun Zhao Lian Tang Xia Faculty of Engineering China University of Geasciences Wuhan Hubei 430074 China) Abstract:There are much weak weathered rockmass and topsoil in tunnel slope its'poor stability makes settlement monitoring much more important to decide the stability of tunnel entrance.Therefore the settlement deformation prediction in tunnel slope is necessary.In view of strong non-linear characteristics of the slope settlement deformation this paper selects BP neural network to predict deformation of the slope and verifies the feasibility and then uses the BP neural network to expand the sample of settlement monitoring.Based on the above the new monitoring samples are analyzed by R/S analysis and Hurst index of settlement-time sequence and deformation rate-time series of the overlaying slope is achieved. Then the two Hurst indexes are bined to determine settlement deformation t...