基于NMI-SS-FOA优化极限学习机的隧道变形预测模型  

Prediction model of tunnel deformation based on NMI-SS-FOA optimized extreme learning machine

作　　者：姜平 徐剑波 杨熙 许文军 任若微 罗学东[2] JIANG Ping;XU Jianbo;YANG Xi;XU Wenjun;REN Ruowei;LUO Xuedong(China First Metallurgical Group Co.,Ltd.,Wuhan 430081,China;Faculty of Engineering,China University of Geosciences(Wuhan),Wuhan 430074,China)

机构地区：[1]中国一冶集团有限公司,湖北武汉430081 [2]中国地质大学(武汉)工程学院,湖北武汉430074

出　　处：《安全与环境工程》2024年第6期48-56,共9页Safety and Environmental Engineering

基　　金：国家自然科学基金项目(42072309)。

摘　　要：为确保隧道施工在安全、经济和高效方面的要求得以满足,对隧道变形进行准确预测是十分必要的。针对隧道变形数据具有非线性和时序性等特征,提出了一种基于NMI-SS-FOA优化极限学习机(ELM)的隧道变形预测模型。该模型首先通过NMI(归一化互信息)法筛选出影响隧道变形的关键参数,将筛选后的参数作为数据集,然后基于扇区搜索机制的果蝇优化算法优化下的极限学习机(SS-FOA-ELM)来预测隧道的变形,并与数学统计预测模型、BP神经网络模型、随机森林方法、SVR(支持向量回归)模型和ELM模型的预测结果进行对比。结果表明:NMI-SS-FOA优化极限学习机的隧道变形预测模型能有效预测隧道变形,其预测结果对应的均方根误差(E_(RMSE))、平均绝对百分比误差(EMAPE)、a_(10)指数(a_(10))和决定系数(R^(2))分别为5.06、19.42%、0.932和0.607,其预测效果较其他预测模型更好;覆盖层厚度(H)、岩体黏聚力(C_(rm))和岩体内摩擦角(φ_(rm))对预测结果影响较大。研究结果可以为隧道施工引起的隧道变形预测和控制提供参考。To ensure that tunnel construction meets the requirements of safety,cost-effectiveness,and efficiency,accurate prediction of tunnel deformation is necessary.Considering the nonlinearity and time-series characteristics of tunnel deformation data,this paper proposes a tunnel deformation prediction model based on NMI-SS-FOA optimized extreme learning machine(ELM).The model first used the normalized mutual information(NMI)method to screen key parameters affecting tunnel deformation,and then used the SS-FOAELM which is optimized by the fruit fly algorithm with sector search mechanism to predict tunnel deformation,comparing the prediction results with those of statistical prediction methods random forest method,BP neural network model,ELM model,and support vector regression(SVR)model.The research results show that the tunnel deformation prediction model based on NMI-SS-FOA optimized ELM can effectively predict tunnel deformation.The corresponding root mean square error(E_(RMSE)),mean absolute percentage error(E_(RMSE)),a_(10) index(a_(10)),and coefficient of determination(R^(2))are 5.06,19.42%,0.932,and 0.607,respectively,indicating better prediction performance than other models.The thickness of the covering layer(H),the cohesion of the rock mass(C_(rm)),and the internal friction angle of the rock mass(ϕ_(rm))have a significant impact on the prediction results.The research results can provide reference for the prediction and control of tunnel deformation caused by tunnel construction.

关 键 词：隧道工程 变形预测 优化算法 极限学习机 归一化互信息法 

分 类 号：X948[环境科学与工程—安全科学] U458[建筑科学—桥梁与隧道工程]