基于智能算法的盾构施工地表沉降预测研究

发布时间：2018-05-22 20:58

本文选题：BP神经网络 + 小波神经网络　；参考：《石家庄铁道大学》2015年硕士论文

【摘要】：盾构法隧道施工凭借其自动化程度高、对环境影响小等特点逐渐成为主流的地铁隧道施工方式。由盾构施工引起的地表沉降对周围的建筑物、地下管线等影响较大,如不对其进行有效控制,可能会引起重大的安全事故。本文首先介绍了盾构施工地表沉降预测在国内外的研究概况,详细叙述了土压平衡盾构机的结构和工作原理,介绍了北京地铁6号线二期工程东小营站~东部新城站为工程概况,研究了地表沉降的机理和发展历程。针对研究对象,本文选用BP神经网络和小波神经网络技术进行研究。在充分考虑盾构施工地表沉降的机理基础上,选取了对地表沉降较为敏感的参数,建立神经网络预测模型。通过对施工现场获得了较多的监测数据进行整理,最终选取一定数量的实验和预测样本,并使用已经建立的神经网络模型进行沉降预测研究。为了达到较好的参数优化目的,本文首先选用蚁群算法,但优化效果易受到蚁群算法收敛速度慢、搜索停滞等缺陷影响,故使用一种收敛速度和稳定性都较好的差分进化算法与其结合来增强算法的性能。通过建立差分进化蚁群神经网络模型,分别对BP神经网络和小波神经网络的初始权值、阈值(伸缩参数和平移参数)进行优化,同时对比了差分进化蚁群算法相对于基本蚁群算法在收敛速度和求解精度方面的优势。论文在最后使用优化后的BP神经网络和小波神经网络分别进行沉降预测,对比了两种模型在收敛速度和预测精度等方面的差异。结果显示,两种模型的预测结果均在工程实际允许范围内,具有较高的参考价值。在实际应用中,应针对各自特点,根据工程具体情况进行选取。论文的研究工作对隧道盾构施工的质量和进度提高均具有较高的应用价值。
[Abstract]:Due to its high degree of automation and small impact on the environment, shield tunneling has gradually become the mainstream subway tunnel construction method. The ground subsidence caused by shield tunneling has a great influence on the surrounding buildings and underground pipelines. If it is not controlled effectively, it may cause serious safety accidents. This paper first introduces the research situation of ground subsidence prediction in shield construction at home and abroad, and describes the structure and working principle of earth pressure balance shield machine in detail. This paper introduces the general situation of Dongxiaoying station ~ east Xincheng station in the second phase of Beijing Metro Line 6, and studies the mechanism and development course of surface subsidence. In this paper, BP neural network and wavelet neural network are selected for research. On the basis of fully considering the mechanism of ground subsidence in shield construction, the more sensitive parameters for surface settlement are selected, and the neural network prediction model is established. More monitoring data were obtained from the construction site, and a certain number of experimental and prediction samples were selected, and the established neural network model was used to predict the settlement. In order to achieve a better goal of parameter optimization, ant colony algorithm is first selected in this paper, but the optimization effect is easily affected by the slow convergence speed of ant colony algorithm, the stagnation of search and so on. Therefore, a differential evolutionary algorithm with good convergence rate and stability is combined to enhance the performance of the algorithm. By establishing a differential evolution ant colony neural network model, the initial weights, thresholds (telescopic parameters and translation parameters) of BP neural network and wavelet neural network are optimized, respectively. At the same time, the difference evolution ant colony algorithm is compared with the basic ant colony algorithm in terms of convergence speed and accuracy. Finally, the optimized BP neural network and wavelet neural network are used to predict the settlement, and the differences of convergence speed and prediction accuracy between the two models are compared. The results show that the predicted results of the two models are within the scope of engineering practice and have a high reference value. In the practical application, according to their characteristics, according to the specific conditions of the project to select. The research work of this paper has high application value to improve the quality and progress of shield tunneling construction.
【学位授予单位】：石家庄铁道大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U455.43

【参考文献】