【作者】 易志强；

【导师】 凌同华；

【作者基本信息】 长沙理工大学 ， 建筑与土木工程， 2011， 硕士

【摘要】 随着隧道工程施工技术和设计理论的快速发展,公路隧道的修建也越来越多。在目前公路隧道的施工中,新奥法的应用相当广泛,隧道监控量测作为新奥法的重要组成部分,已逐步成为隧道工程设计的项目和确保工程施工安全的重要手段,且监测技术也从中得到了飞跃性的发展,并直接影响到隧道的结构形式、施工方法、支护参数、造价、工期等。但由于监测手段以及后期数据处理分析方法的不完善,信息反馈不够及时等问题,致使监测数据在验证事前设计和指导事后施工方面还远未发挥其应有的作用。针对上述问题,本文作者提出了隧道监控量测数据挖掘及隧道变形预测的研究课题,并基于大量的现场监测数据和地质资料,在已有研究成果基础上进行了深入系统的研究。首先,针对现场所得监测数据存在杂乱性、重复性、不完整性等问题,采用了数据挖掘理论中的数据预处理系统分别对监测数据进行了数据清理、变换、集成、规约等处理,大大提高了数据挖掘模式的质量,降低了后续挖掘所需时间。其次,采用SPSS软件在依据相关回归分析理论及相应隧道变形模型基础上对经过预处理后的不同围岩断面的地表沉降、拱顶下沉和周边收敛等监测数据进行回归分析,验证了SPSS软件对庞大数据进行多元回归分析的有效性。根据所得结果,可以初步计算隧道开挖后变形的极限值以及其达到稳定状态所需时间,为预留变形量的调整及后期工序的变更提供有力依据。再次,针对影响隧道围岩变形的因素具有复杂性、随机性、模糊性等特点,首次应用自适应神经模糊推理系统(ANFIS),并基于Takagi-Sugeno(T-S)模型建立了隧道位移的预测模型。该预测模型具有收敛速度快、稳定性好、训练过程可重复、预测精度高等特性,可有效地预测不同围岩下隧道变形的一般规律。本文所做研究工作,立足于学科前沿,采用最新数学计算方法和手段对隧道监控量测数据挖掘及隧道变形预测进行了研究,具有较高的理论和应用价值,为隧道监控量测数据的处理提供了一种新的有效分析手段。更多还原

【Abstract】 With the rapid development of construction technology and design theory of tunnel engineering, more and more highway tunnels are constructed. New Austrian Tunneling Method (NATM) is widely used in the current construction of highway tunnels, and the monitor measuring of highway tunnel has gradually become a project of tunnel engineering design and an important mean to ensure the safety of construction as an important component of NATM. Monitoring technology has also been developed quickly, and it has a direct impact on the tunnel structure, construction method, supporting parameters, cost, duration, etc. Because of the imperfect monitoring tools , post-processing analysis of data, information feedback not in time and other problems , the monitoring data can not play its due role in the verification aspects of pre-design and follow-up construction. According to the mentioned problems above, the author put forward the research subject of monitoring measurement data mining and deformation prediction of a tunnel, and conducted the thorough system by using the existing research results based on the on-site monitoring data and geological data in the study.First of all, according to the characteristics(clutter, repeatability, and no-integrity) of the monitoring data, data cleaning, transformation, integration, protocols and other processing were carried out by the data pre-processing system of data mining. The process has greatly improved the quality of data mining models and reduced the time required for subsequent mining.Secondly, according to regression theory and the corresponding deformation model of a tunnel, the surface subsidence of soil section, the crown settlement and the surrounding convergence of different surrounding rock sections are analysed by using regression analysis through SPSS software ,and it verified the effectiveness of using multiple regression analysis to process huge data in SPSS software. On the basis of the results,we can preliminary calculate the limit of deformation after the tunnel was excavated and the time needed for stabilization state,and it also can provide a strong basis for the deformation adjustment and post-processing arrangement.Lastly, for the influence factors (Complexity, randomness, fuzziness) of the surrounding rock deformation displacement, based on the Takagi-Sugeno(t-s) model, adaptive neuro-fuzzy inference system(ANFIS) is adopted in the first time to establish a prediction model of the tunnel displacement. Because the fast convergence rate, good stability, repeatability of the training process, and high prediction accuracy of this forecast model, the general rule of tunnel deformation in different surrounding rock can predicted effectively.The research work in this paper, based on forefront subject, adopted the latest mathematics calculation method and means to study the monitoring measurement data mining and deformation prediction of a tunnel. It is valuable both in theories and the applications for providing a new effective method to process tunnel monitoring measurement data.更多还原