摘  要

目前，中国己成为世界上地铁施工领域中使用盾构最多的国家之一。近年来，以盾构施工过程中产生不允许的土层变形位移和过大的地表沉降所引发城市环境土工安全事故呈明显上升趋势，如何掌握复杂条件下城市地铁盾构施工过程中地表变形的时空演化机理和规律，进而准确预测其时空演化过程，并实现地表变形预警系统，已成为盾构隧道现代化建设中的一个函待解决的重要课题。

本文借鉴国内外相关研究，针对武汉市轨道交通二号线越江地铁盾构隧道工程，以时间序列分析和随机介质理论为基础，综合运用岩土几工程反分析、人工智能、系统辨识、信息融合等方法，系统分析了盾构隧道施工引起的地表变形时空演化规律，成功实现了盾构施工引起地表变形的智能预测和预警。

首先，本文以武汉市轨道交通二号线越江地铁盾构隧道工程为背景，分析指出越江地铁隧道盾构施工风险均与地表变形的程度直接或间接有关，控制越江地铁隧道盾构施工风险的重点和关键之一就是预测并控制陆地段和江中段地铁盾构施工引起的地表变形的程度，降低由地表变形导致一系列次生灾害的可能。

盾构施工地表变形时间序列具有非等间距、小样本、趋势性和自相关性等关键特征，本文提出基于Weibull一ARIMA的盾构施工地表变形时间过程模型，并完整的给出了建模流程和算法。该模型能够很好的反向解构盾构施工引起的地表变形时间序列，特别是在拟合精度和建模效率上，均明显优于单纯的weibull模型或者ARIMA模型。通过该模型的应用，可以快速准确的挖掘盾构施工地表变形时间序列蕴含的时间效应特征参数，实现对盾构施工地表变形时间过程的描述和刻画。

盾构施工引起的地表变形的三维空间位移场也就是变形空间分布的范围和大小。本文运用改进的随机介质理论给出了由盾构施工引起的地表变形空间分布计算过程。在此基础上，提出基于多重自适应变异粒子群优化(MAMPSO)算法实现盾构地表变形空间分布特征参数的反分析。该算法与传统的Powell算法、基本PSO算法相比，在盾构施工地表变形空间分布特征参数反演中具有较高的识别性能和精度，为分析盾构施工地表变形空间分布规律提供了新的方法。盾构施工过程是一个随时间和空间不断演化的复杂系统。本文提出了后构施工地表变形时空演化系统建模思路，建立了基于MAMPSO一RBFNN的后构施[地表变形时空演化智能预测模型，并转化为RBF神经网络模型结构及参数的非线性函数优化问题，利用MAMPSO算法优化并实现了盾构施工地表变形时空演化过程的实时智能预测。

盾构施工地表变形预警系统是信息技术在地下工程中的重要应用之一。本文构建了基于多源信息融合的盾构施工地表变形预警模型，分别实现了盾构施工地表变形警兆融合、区域和工点预警决策融合，为施工现场提供了具有警情识别、警情分析、警情预测、警情评价、警情决策与一体的预警系统。

关键词:地铁工程   泥水盾构隧道施工    地表变形时空演化    随机介质理论      MAMPSO算法     RBF神经网络     时间序列分析    多源信息融合

Abstract

Nowadays, Chinahas become one of the largest shield tunnel markets in metro construction inthe world. In recent years, it is obviously that the urban environmental geotechnicalaccidents happened more and more which caused by unallowed stratum deformationand ground surface settlements during the shield tunneling.

During the metroshield tunnel constructions under complex conditions, how to explain the time-spaceevolution mechanism and laws of the ground surface

settlements, and then predicttheir time-space evolution process and achieve early warning system for them .It has become an important issue waiting for solving

in modern shield tunnelconstruction.

Profited fromthe related domestic and foreign research results and based on time seriesanalysis and stochastic medium theory, this thesis proposes to use

backanalysis of geotechnical engineering, artificial intelligence, system identificationmethods to explain the time-space evolution mechanism and laws of the

groundsurface settlements, and achieve prediction and visualizational early warning forthem in Wuhan Metro Line II Yangzi river-crossing shield tunnel construction.

Firstly, therelationship between construction safety risks and ground surface settlementsin river-crossing metro shield tunnel is analyzed. This thesis point

out thatthe focus and key to control construction safety risks and the accidents isprediction and controlling of the ground surface settlements in the land andriver

section during river-crossing metro shield tunnel construction.

Ground surfacesettlement time series have key characteristics such as unequally space, smallsample size, trend and autocorrelation. Ground surface settlement

time series modelbased on Weibull-ARIMA and its complete modeling process and algorithms are proposed.And this model can explain the ground surface

settlement time series in shield tunnelconstruction more efficiently and accurately than the Weibull model, or ARIMA models.The applications of this model can gain the time characteristic parameters inthe ground surface settlement time series fast and accurately and fulfill theirdescription and characterization.

The three-dimensional displacement field is the scope and size of the spatial distributionof the ground surface settlement. In this thesis, the spatial distribution

calculationof the ground surface settlements is presented using the improved stochastic mediumtheory. Additionally, multi-adaptive mutation particle swarm

optimization (MAMPSO)algorithm is proposed for back analysis of the spatial distribution characteristicparameters in the ground surface settlements. MAMPSO

algorithm has a highidentification performance and accuracy with the traditional Powell algorithmand the basic PSO algorithm and provides a new method to analyze the spatialdistribution law of the ground surface settlements.

Shield tunnelconstruction process is a complex evolving system with time and space.Theprinciple of modeling the time and space evolution system of the ground surfacesettlements during shield tunnel construction is illustrated in this thesis.Based on the MAMPSO, this paper proposes the Radial Basis Function NeutralNetworks to establish the intelligent prediction model of the ground surfacesettlements. The modeling is equal to the nonlinear function optimizationproblem of the

structure and parameters in RBFNN, which can be solved byMAMPSO.

One of the mostimportant applications of information technology in underground construction isthe early warning system of the ground surface settlements during shield tunnelconstruction. Based on multi-source information fusion theory, the earlywarning model of the ground surface settlements engineering is proposed.

The thesis use the model and other information technology to achieve the functionssuch as feature fusion and decision fusion for local and integrated earlywarning,

provides early warning system which including waring identification,analysis, prediction, accessment and decision for the ground surface settlementduring shield

tunnel construction.

Key words: metro construction     slurryshield tunneling    ground surfacesettlement timespace evolution time series analyses     stochastic medium theory     multiple adaptive mutation PSO    RBFneutral network     multi-sourceinformation fusion  ationfusion