广州地铁盾构隧道施工中地表沉降的SVR法预测研究

发布时间：2018-03-25 08:50

本文选题：广州地铁　切入点：盾构机　出处：《暨南大学》2016年硕士论文

【摘要】：随着城市地下铁路建设的快速发展,盾构隧道施工技术由于其良好的地层适应性和安全性等优点城市地铁建设中得到广泛的应用。在实际隧道工程中,盾构施工引起的地表沉降往往容易引发相关安全问题,因此,地表沉降预测和控制成为了工程界最为关注的问题之一。针对该问题,以广州地铁十三号线新塘~官湖区段工程为背景,结合时下热门的大数据理念,选取合适的机器学习方法对工程相关数据进行处理,从而实现对地表沉降预测的研究。研究内容主要包括:1、将盾构施工参数分为盾构机掘进参数和水文地质参数两类,利用实际工程数据研究讨论了不同参数对地表沉降的影响,总结了影响地表沉降的主要因素。2、利用已有的经验公式和理论公式分别对实际工程数据对进行拟合对比,证明了Peck公式更适用于实际工程沉降预测,并统计出该区段沉降横向分布规律。3、针对传统神经网络容易过拟合和“维数爆炸”的缺点,创新性地采取支持向量回归机(SVR)算法对Peck公式中的地表中心沉降值参数进行预测改进。通过严谨的理论推导阐明SVR算法原理,并利用其对地表沉降进行训练并预测。引入相对误差和均方误差(MSE)作为模型的评估标准,对比不同参数对于模型的影响,结果表明盾构掘进参数对地表沉降具有较大的影响。4、最后通过与传统BP神经网络方法进行对比来验证该方法的有效性,结果表明SVR模型比BP具有更好的估计能力。
[Abstract]:With the rapid development of urban underground railway construction, shield tunnel construction technology has been widely used in urban subway construction because of its good stratum adaptability and safety. The ground subsidence caused by shield tunneling is often easy to lead to safety problems. Therefore, the prediction and control of surface settlement has become one of the most concerned problems in the engineering field. Based on the project of Xintang to Guanghu section of Guangzhou Metro Line 13, combined with the popular concept of big data, we select the appropriate machine learning method to deal with the relevant data of the project. The main contents of the research include: 1, the shield construction parameters are divided into shield tunneling parameters and hydrogeological parameters, and the influence of different parameters on surface subsidence is discussed by using the actual engineering data. This paper summarizes the main factors that affect the surface settlement. The empirical formula and the theoretical formula are used to fit and contrast the actual engineering data respectively. It is proved that the Peck formula is more suitable for the prediction of the actual engineering settlement. The lateral distribution rule of settlement in this section. 3. Aiming at the disadvantages of the traditional neural network, it is easy to over-fit and "dimension explosion". In this paper, the support vector regression (SVR) algorithm is used to predict the surface center settlement parameters in the Peck formula. The principle of the SVR algorithm is explained through rigorous theoretical derivation. The relative error and mean square error (MSE) are introduced as the evaluation criteria of the model, and the effects of different parameters on the model are compared. The results show that the shield tunneling parameters have a great influence on the surface subsidence. Finally, the effectiveness of the method is verified by comparing it with the traditional BP neural network method. The results show that the SVR model has better estimation ability than BP.
【学位授予单位】：暨南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U455.43;U231.3

【相似文献】