隧道围岩大变形高斯过程回归预测模型及其工程应用  

Gaussian Process Regression Prediction Model for the Big Deformation of the Tunnel Rock and Its Application

作　　者：张云鹏[1] 李利平[1] 贺鹏[1] 秦承帅 

机构地区：[1]山东大学岩土与结构工程研究中心,济南250061

出　　处：《科学技术与工程》2018年第1期122-127,共6页Science Technology and Engineering

摘　　要：为解决围岩大变形中出现的高度非线性时序问题,引入可处理小样本、高度非线性问题的高斯过程回归(GPR)理论。以日本长崎嬉野隧道及四川地区的王登隧道两个典型大变形隧道工程为例,以长期监控量测所得数据为训练样本,运用GPR方法进行构建隧道围岩大变形预测模型。得出主要结论:(1)相较于ANN、SVM、GM(1,1)等常规预测模型,GPR预测模型可更为精确地对隧道拱顶沉降值与净空收敛值进行预测;(2)对岩体出现大变形拐点的非线性阶段,其预测趋势更符合工程实际,可更为合理的对围岩大变形初期阶段实现预测,进而及早地实现对施工预警以及支护方案的比选优化;(3)GPR算法经工程案例验证,能高效精准地对隧道围岩变形进行预测,可为类似工程提供借鉴。To solve the problem in the large deformation of surrounding rock of highly nonlinear time sequence problem,the Gaussian process regression( GPR) theory is introduced,which can handle small samples and highly nonlinear problems. Two tunnels are selected as typical samples: Japan Nagasaki Ureshino Tunnel and Sichuan Wangdeng Tunnel,for the long-term monitoring data from the training sample,using GPR method to build the tunnel surrounding rock large deformation forecast model. Main conclusions:(1) compared with the traditional forecasting models such as ANN,SVM,GM( 1,1),GPR prediction model can be more accurate to predict tunnel vault subsidence value and the net convergence value;(2) on a large rock mass deformation inflection point of nonlinear phase,its forecast trend more accord with the engineering practice,can be more reasonable for large deformation of surrounding rock of the early stage of implementation,and then as soon as possible to realize the construction of early warning and support scheme comparison optimization;(3) GPR algorithm is verified by engineering examples,can be efficient and accurate to predict the deformation of surrounding rock of the tunnel,can provide reference for similar projects.

关 键 词：隧道工程 高斯过程回归 变形预测 围岩 

分 类 号：U456.3[建筑科学—桥梁与隧道工程]