基于DEACO-WNN的盾构施工地表沉降预测  

Study on Predicting the Surface Settlement for Shield Tunneling Based on DEACO-WNN

作　　者：郝如江[1] 季雁鹏 倪振利 

机构地区：[1]石家庄铁道大学,石家庄050043 [2]中铁二十二局集团第一工程有限公司,北京100043

出　　处：《铁道工程学报》2015年第1期12-16,共5页Journal of Railway Engineering Society

基　　金：国家自然科学基金(51375319);河北省杰出青年科学基金(E2013210113);河北省百名优秀创新人才支持计划(BR2-222)资助

摘　　要：研究目的:近年来,我国地铁建设飞速发展,全国有近四十个城市在修建地铁。盾构施工是地铁建设的主要施工方式,由于盾构施工引起的地表沉降问题,为施工安全带来很多的困扰。为了减少施工过程中的安全隐患,提高施工质量和效益,对盾构施工引起的地表沉降进行研究意义重大。研究结论:本文通过对盾构施工地表沉降的影响因素和预测模型进行研究,得到以下结论:(1)在研究了对于盾构施工地表沉降较为敏感的参数中选取了覆土厚度、压缩模量、凝聚力、天然密度、内摩擦角、千斤顶推力、注浆压力,并将其作为输入层建立了小波神经网络预测模型;(2)用差分进化蚁群算法以相对熵为标准对小波神经网络的初始权值、伸缩参数和平移参数进行优化,使得预测模型的收敛速度和预测精度都有显著提高;(3)通过北京地铁6号线实地工程数据进行验证,预测模型的最小误差仅为0.5%,预测精度在实际工程所允许的范围内;(4)本文所建立的模型对盾构施工安全防护领域具有指导意义,通过对地表沉降的有效预测来改进施工工艺,进而增强施工的安全性。Research purposes： In recent years, the subway career is developing rapidly, a number of new lines were built. Surface settlement caused by security risks during the construction significant. shield tunneling brings more trouble for construction safety. A research on reduce and improve construction quality of surface settlement caused by shield tunneling isResearch conclusions：In this paper, the factors of surface settlement caused by shield tunneling and prediction model were studied, the following conclusions were got ： （ 1 ） A wavelet neural network （WNN） prediction model was created by soil covering thickness, compression modulus, cohesion, natural density, angle of intemal friction, jack thrust, pressure grouting. （2）The surface settlement model was created by the wavelet neural network（ WNN）, which its original weight, scaling parameters and translation parameters values were obtained by the differential evolution and ant colony optimization algorithm （DEACO） and the goal of optimization was relative entropy. The training speed and prediction accuracy was improved greatly using the DEACO - WNN. （3） The measured data in Beijing 6th subway line were tested in the presented model, and the minimum prediction error was just only 0.5%, the prediction accuracy within theallowed range. （4）The model created in this paper has a good guidance for the security protection area of shield tunneling, it could be helpful to improve the construction technique through effective prediction of surface settlement, and enhancing the safety of construction.

关 键 词：小波神经网络 差分进化蚁群 相对熵 盾构施工 沉降预测 

分 类 号：U455[建筑科学—桥梁与隧道工程]