摘  要

大规模的地铁建设缓解了城市交通压力，却激发了城市岩土工程的安全风险。岩土工程问题的复杂性促使工程领域迫切的需要一套系统化的方法积累工程经验、探索工程规律、利用工程知识。本文遵循知识管理的技术路线，针对地铁隧道工程领域最受人关注的地表沉降指标，提出了一套从数据到智慧的系统化研究方法，该方法既有助于工程经验的积累，也为现场安全决策提供支持。

首先，本文构建了面向隧道施工地表沉降分析的时空情境多维数据仓库，将零散的多源异构数据整合成集成情境的价值信息。在从数据到信息的转换过程中，本文创新性地提出了地表沉降增长的Logistic 模型；基于该模型，本文运用非线性回归及随机借补技术解决了沉降监测数据不等距缺失问题；本文还运用离散小波变换技术滤除了日沉降中的监测误差，综合提升了数据的质量和信息的价值。

基于上文构建的时空情境多维数据仓库，本文进一步从时间、空间、情境等多个视角挖掘了地表沉降规律，提炼了能够指示地表沉降安全风险的相关指标。在从信息到知识的转换过程中，本文创新性地提出了双隧道沉降槽模型，并运用Levenberg-Marquardt 方法解决了该模型的非线性拟合问题；基于大量的拟合结果，本文发现了沉降槽宽度的时不变性及其与地质水文分布的相关性；本文还提出了地层损失的计算方法，得出了地层损失的分布情况和增长规律。

最后，综合运用前文得到的地表沉降时、空、情境规律，本文建立了地表沉降安全风险预警机制以指导工程实践。在这一从知识到智慧的转换过程中，本文创新性地提出了地表沉降风险分级区域的划分方法，该方法由高斯分布经验法则推广而来，并结合前文得到的沉降槽宽度预测公式可对地表沉降安全风险进行预评估；本文还引入统计过程控制（SPC）思想提出了对地表沉降、地层损失等风险指标进行异常数据识别的方法。

关键词：隧道施工；地表沉降；数据仓库；数据挖掘；非线性拟合

Abstract

Largescale construction of subway system can ease the pressure on urban traffic, butinspire the geotechnical safety risk in urban area. There is an emergent need for asystematic way to accumulate engineering experience, to explore engineering rules, to implement engineering knowledge, because the geotechnical works are very complex.This dissertation follows the route of knowledge management technology and proposes a systematic method named ‘from data to wisdom’ to evaluate tunneling induced ground settlement which deserves most safety attention.This method not only contributes to the accumulation of experience in engineering, but also supports safety decision on site.

First, a multidimensional data warehouse is established to facilitate the analysis of ground settlement and integrate the fragmented and heterogeneous data into valuable information. In the conversion process of data to information, a Logistic model is proposed to describe the growing pattern of ground settlement and solve the data missing problem by using nonlinear regression and random imputation techniques.wavelet analysis method is also implemented to filter out the measurement error and enhance the quality of the data and the value of information.

Then, the rules of ground settlement are explored in multiple perspectives and some safety indicators are refined by using the data warehouse.In the conversion process of information to knowledge, the settlement trough model of twintunnel is proposed and LevenbergMarquardt method is implemented to solve the problem of nonlinear model fitting; the fitting results show the time invariance of the settlement trough width and its correlation with the distribution of the geological and hydrological condition; this dissertation also proposes the calculation method of ground loss and concludes its distribution and growth pattern.

Finally, an early warning mechanism for ground settlement control is established to guide the engineering practice by comprehensively applying the proposed rules of ground settlement. In the conversion process from knowledge to wisdom, a method to scope the safety area is proposed by extending the rule of Gaussian distribution and adopting the proposed prediction method of the width of ground settlement trough; the statistical process control (SPC) method is also implemented to identify the outlier of ground settlement, ground loss and etc.

Keywords: Tunnelling construction; Ground settlement; Data warehouse; Data mining;Nonlinear fitting