基于Logistic映射GWO的盾构姿态预测  

Prediction of shield posture using logistic grey wolf optimizer algorithm

作　　者：刘云花 张亚辉 LIU Yunhua;ZHANG Yahui(CCCC South China Engineering Bureau Co.,Ltd.,Changsha 410000,China;Department of Civil Engineering,Shijiazhuang Tiedao University,Shijiazhuang 050043,China;Hebei Urban&Rural Construction School,Shijiazhuang 050031,China)

机构地区：[1]中交中南工程局有限公司,长沙410000 [2]石家庄铁道大学土木工程学院,石家庄050043 [3]河北城乡建设学校,石家庄050031

出　　处：《交通科技与经济》2025年第2期49-56,共8页Technology & Economy in Areas of Communications

基　　金：中交第一公路工程局有限公司科技研发项目(KJHT2018-22)。

摘　　要：针对地铁隧道盾构姿态难以控制的问题,提出基于Logistic映射灰狼算法(GWO)的盾构姿态预测方法。该方法引入Logistic映射生成初始值,并针对灰狼算法中的收敛因子A进行分解,分为决策因子A 1与衰减因子A 2,以改善收敛因子在全局与局部搜索时的能力不足问题,可以有效学习盾构时序数据中的长期依赖关系。研究依托武汉地铁11号线东段马房山站~虎泉站数据集对所提方法进行验证。以盾尾水平偏移预测为例,L-GWO模型预测的相关系数(R)达到了0.9985,均方根误差(RMAE)为1.23 mm,相较于GWO模型和粒子群,L-GWO模型可以更准确地预测盾构姿态。此外,与GWO、GA-GWO、PSO-GWO模型相比,Logistic映射在防止网络退化、加强盾构数据特征学习能力方面的作用显著。Aiming at the problem of difficult control of shield tunnel attitude in subway tunnels,a shield posture prediction method based on Logistic Mapping Grey Wolf Algorithm(GWO)is proposed.This method introduces logistic mapping to generate initial values and decomposes the convergence factor A in the grey wolf algorithm into decision factor A 1 and attenuation factor A 2 to improve the insufficient ability of the convergence factor in global and local searches.It can effectively learn long-term dependencies in shield time-series data.Based on the dataset from Mafangshan Station to Huquan Station on the eastern section of Wuhan Metro Line 11,the proposed method was validated.Taking the prediction of shield tail horizontal offset as an example,the correlation coefficient(R)predicted by the L-GWO model reached 0.9985,and the root mean square error(RMAE)was 1.23 mm.Compared with the GWO model and particle swarm optimization,the L-GWO model can more accurately predict the shield posture.In addition,compared with the GWO model GA-GWO model and PSO-GWO model,logistic mapping plays a significant role in preventing network degradation and enhancing the feature learning ability of shield tunneling data.

关 键 词：隧道工程 盾构隧道 机器学习 灰狼算法 混沌映射 姿态预测 

分 类 号：U459[建筑科学—桥梁与隧道工程]