基于小波和时间序列分析组合模型的地铁隧道变形预测研究

发布时间：2018-03-22 11:42

本文选题：地铁隧道变形监测　切入点：小波去噪　出处：《南京师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：目前我国各大城市均在建设高效的地铁隧道网,地铁隧道在施工和运行中由于受多种因素影响会产生变形,变形如果超出安全范围将引起严重后果,所以建立及时有效的预报模型具有重要的意义,但各个模型都有局限性,单一的模型往往预测精度较低,因此需要将现有模型有针对性的组合和优化。地铁隧道变形数据具有动态、平稳、含噪声的特点,时间序列分析在处理预测动态平稳信号时有很好的效果,而小波分析能够作为预处理工具,有效消去原始信号中的噪声部分,从而提高预测精度。本文提出小波和时间序列分析组合模型,对地铁隧道变形进行预测。论文主要的研究内容如下:(1)地铁隧道变形预测方法和变形情况分析研究地铁变形预测方法,从理论基础、分析方法、数据量要求和研究重点等方面对常用方法特点进行分析与比较;研究地铁隧道结构沉降的影响因素,基准网布设方法、测量技术要求等内容;以南京地铁十号线隧道结构沉降数据为例,分析地铁隧道单点变形数据特征以及全线监测点、车站主体结构及各个区间的变形情况。(2)小波分析和时间序列分析模型研究研究小波变换和阈值去噪的基本理论,通过改变小波函数和阈值估计方法进行去噪效果比较,选取适合本文数据的小波函数以及阈值估计方法;研究时间序列分析的分类、特点以及AR、MA、ARMA模型基本原理,重点研究模型识别、定阶和参数估计方法,利用地铁隧道变形数据进行单一时间序列分析建模和预测。(3)组合模型的构建和实例验证结合小波和时间序列分析模型的特点,通过两种不同组合方式分别进行拟合预测并与单一模型拟合预测结果进行比较,验证组合模型由于去除了原始信号中的噪声,信号变的更加平滑,使时间序列分析充分发挥它在处理平稳信号时的优势,取得更好的拟合效果;通过评价指标验证对去噪后的分量进行时序预测再重构的组合模型,拟合准确度和预测精度更高;最后利用效果更好的组合方式对变形突出的中胜站、龙华路站以及中胜—元通区间的沉降量进行预测,研究变形趋势并分析变形原因,有利于及时发现问题并采取相应措施。
[Abstract]:At present, all the major cities in our country are building an efficient subway tunnel network. The subway tunnel will be deformed due to the influence of many factors in its construction and operation. If the deformation exceeds the safe range, it will cause serious consequences. Therefore, it is of great significance to establish a timely and effective forecasting model, but each model has its limitations, and a single model often has low prediction accuracy. Therefore, it is necessary to combine and optimize the existing models. The deformation data of subway tunnel have the characteristics of dynamic, steady and noisy. The time series analysis has a good effect in the prediction of dynamic stationary signals. Wavelet analysis can be used as a preprocessing tool to effectively eliminate the noise part of the original signal, thus improving the prediction accuracy. In this paper, a combined model of wavelet and time series analysis is proposed. The main contents of this paper are as follows: 1) the method of subway tunnel deformation prediction and the analysis of subway deformation. This paper analyzes and compares the characteristics of common methods from the aspects of data requirement and research emphasis, studies the factors affecting the settlement of subway tunnel structure, the method of setting up the benchmark network, and the technical requirements of measurement, and so on. Taking the settlement data of the tunnel structure of Nanjing Metro Line 10 as an example, the characteristics of the single point deformation data of the subway tunnel and the monitoring points of the whole line are analyzed. Wavelet analysis and time series analysis model study the basic theory of wavelet transform and threshold denoising, and compare the denoising effect by changing wavelet function and threshold estimation method. The wavelet function and threshold estimation method suitable for the data of this paper are selected, the classification and characteristics of time series analysis and the basic principle of ARMA-ARMA model are studied, and the methods of model identification, order determination and parameter estimation are emphatically studied. Using the deformation data of subway tunnel to model and predict the single time series analysis. The construction of the combined model and the verification of the example show the characteristics of the wavelet and time series analysis model. The results of fitting and forecasting by two different combinations are compared with the results of single model. It is verified that the combined model is smoother because of removing the noise from the original signal. Make the time series analysis give full play to its advantages in processing stationary signals, obtain better fitting effect, verify the combination model of the de-noised components for time series prediction and re-reconstruct through the evaluation index, the fitting accuracy and prediction accuracy are higher. Finally, the settlement of Zhongsheng Station, Longhua Road Station and Zhongsheng-Yuantong Section with outstanding deformation are forecasted by better combination method, and the trend of deformation is studied and the cause of deformation is analyzed, which is helpful to find the problem and take corresponding measures in time.
【学位授予单位】：南京师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U456.3

【参考文献】