摘 要

当前，我国地铁建设进入发展黄金期。伴随地铁工程的快速发展，施工过程中的安全事故却时有发生，严重威胁着周围环境及居民的人身财产安全。地铁基坑安全事故频发不仅是因为基坑工程“大、紧、深”的特点，更是由于风险管理的缺陷，如风险预测不准确、风险规律把握不准、风险控制针对性不强、风险管理水平低等。因此，本文针对地铁工程安全风险的预测、风险发生规律及可视化方法等问题进行研究，提出了系统的风险管理方法，并以武汉地铁某车站施工为背景，对提出的方法和模型进行工程验证，结果表明该方法能够准确的预测风险状态，进行风险的致因分析和可视化控制。

首先，针对基坑风险管理现状，本文通过对国内外风险管理、评估方法的研究，重点讨论了三大类风险管理方法（经验法、数值模拟法、机器学习法）的优缺点，本文采用Support Vector Machines(SVM)方法用于基坑施工风险预测，BayesianNetworks(BN)方法用于风险诊断分析，最后基于BuildingInformationModel(BIM)技术，实现风险可视化管理与控制。

其次，基于相关理论，构建本文提出的风险管理方法的基本框架和流程。详细地给出了基于SVM的基坑风险预测步骤，基于BN的风险诊断的分析流程，以及基于BIM的风险可视化模型的控制过程。通过三种方法的结合使用，可以实现事前、事中和事后的全过程风险控制。

最后，以某地铁车站为背景，在风险识别的基础上，利用SVM实现基坑风险等级分类预测，有效地解决了小样本条件下风险预测的准确性问题。继而，根据风险预测结果，模拟基坑总体风险的时空演化规律。同时以SVM预测结果为基础，结合BN方法，进行地下连续墙侧移风险推理，确定了导致地下连续墙侧移的关键因素，进而有针对性地进行风险控制。最后结合BIM技术进行基坑风险控制，风险状态可实时、直观地展示，风险应对措施自动生成，并可对关键部位进行施工模拟。通过实际工程案例，验证了该方法可有效预测地铁基坑施工风险，为地铁车站施工过程安全控制提供有效保障。

关键词：地铁深基坑；风险预测；风险诊断；风险控制；支持向量机；贝叶斯网络；BIM

Abstract

In recent years,the development ofmetroconstructionsinChinahas comeinto the golden period.With the rapid development of themetroconstructions, the safety accident during the construction process has occurredfrequently,which madea serious threat to the surrounding environment and the residents’ property safety.Thereasonoffrequent accidentsisnot only the characteristics of the foundation pit-“big, tight, deep”, but also the shortcomings of risk management, such as inaccurate riskprediction, inaccurate risklaws,nottargeted risk controlandlow risk management level.Therefore, this paper studies the riskprediction, the occurrence law and the visualization methodofsafetyrisks in metroconstructions, the systematic risk management methodwasproposed.The proposed methodwasverifiedby taking the Wuhan Metrostationas an example.The results show that the method can predict the risk statesaccurately, implementanalysis of risk causes and visual control.

Firstly,amingto the current situation of risk management, this paper focusedon the development of risk management methods and evaluation methods at home and abroad(mainlyempirical, numerical simulationmethods,machine learning methods).The advantages and disadvantages of the three types of methodswerediscussed emphatically.Itwasdetermined thatSupportVectorMachine (SVM)wasused for risk prediction, BayesianNetworks(BN)wasused for risk diagnosis, BuildingInformationModel(BIM)wasused for risk visualization managementand control.

Secondly, the basic framework and process of the risk managementwereconstructedbased on the relevant theories. The steps of risk prediction based onSVM, the analysis process of risk diagnosis based on BN, and theflowof risk visible control based on BIMweregiven in detail. Through the combination of the three methods,feed-forward, concurrent and back-forwardrisk controlcould be achieved.

Finally, taking the Wuhan Metrostationas an example, on the basis of risk identification, SVMwasused to predict the risklevels, whichcouldeffectively solve the difficulty in data collection of foundation pit engineering.Then, according to the results of risk prediction, the temporal and spatial evolutionlawsof the overall construction riskinthe metrostationwereobatained.At the same time, based on the prediction results,BN wasused to analyzethe reasons of diaphramwallsway.The key factors leading to thediaphramwallswayare determined,sotheriskcontrolcould beperformed targetly.At last,BIM technologywas introducedfor risk controlinfoundation pit.The risk statescan be displayed in real time, the risk responsereports weregeneratedautomatically, and the keyconstruction processcan be simulatedin advance. Through the actual engineering case, itwasproved thatthemethod can effectively predict the risksinmetroconstructionsand provide effective guarantee for the safety controlinconstruction process.

Key words:deepmetrofoundation pit; risk prediction; risk diagnosis; risk control; support vector machine; BayesianNetworks; BIM