EM算法及其在变形监测数据处理中的应用

发布时间：2018-07-31 06:56

【摘要】：为保证工程建（构）筑物施工过程中的安全，有必要对其进行系统监测。通过监测及利用观测数据资料，分析和把握建（构）筑物的变形特征及变形的发展规律，进而对其变形趋势进行预测和分析。由于人为、仪器等各种影响因素的存在，造成观测数据缺失或者观测值中含有失误差或者反常误差，导致测量数据结果缺失不完全。但当缺失的观测值对于数据处理必不可少时，就有必要对缺失数据进行处理，以使缺失测量数据处理的质量得到有效提高，并进一步提高测量的精度和可靠性。论文根据EM算法的原理，以城市深基坑工程桩顶沉降监测数据为研究对象，展开对缺失监测数据处理的拟合预测研究，主要做了如下的研究： 1.综述现有测量数据处理中常用方法，通过对比多种不完全测量数据处理方法，分析各种方法的适用条件和优缺点。 2.对EM算法的基本原理与性质进行简单介绍，，分析数据的缺失机制与模式，确定本文研究数据的模式，最后并给出数据处理效果判断依据。 3.将EM算法与切比雪夫多项式回归分析法结合应用于不完全测量数据处理中，研究和探讨基于EM算法的多项式回归分析在测量数据处理中的实现步骤，并给出公式推导，为实现基于EM算法的不完全测量数据处理做了理论铺垫。 4.分别运用指数平滑法与BP神经网络模型法对实测沉降数据进行预测，对比EM算法处理结果，发现EM算法的整体预测精度最高。说明在不完全测量数据处理中，EM算法是合理可行的。
[Abstract]:In order to ensure the safety of construction, it is necessary to monitor it systematically. Through monitoring and using the observation data, the deformation characteristics and the development law of the building are analyzed and grasped, and the trend of deformation is predicted and analyzed. Because of the existence of various influence factors, such as human beings and instruments, the missing observation data or the error or abnormal error in the observation value lead to the incomplete result of measurement data. But when the missing observation value is necessary for data processing, it is necessary to process the missing data, so that the quality of missing measurement data processing can be improved effectively, and the accuracy and reliability of measurement can be further improved. According to the principle of EM algorithm, taking the monitoring data of pile top settlement in urban deep foundation pit engineering as the research object, this paper carries out the research of fitting and forecasting the missing monitoring data. The main research work is as follows: 1. This paper summarizes the methods commonly used in measurement data processing, and analyzes the applicable conditions, advantages and disadvantages of these methods by comparing various incomplete measurement data processing methods. This paper introduces the basic principle and nature of EM algorithm, analyzes the missing mechanism and pattern of data, determines the pattern of data research in this paper, and finally gives the basis for judging the effect of data processing. Combining EM algorithm with Chebyshev polynomial regression analysis method in incomplete measurement data processing, the realization steps of EM algorithm based polynomial regression analysis in measurement data processing are studied and discussed, and the formula derivation is given. In order to realize the incomplete measurement data processing based on EM algorithm. 4. The exponential smoothing method and BP neural network model method are used to predict the measured settlement data respectively. Comparing the results of EM algorithm, it is found that the whole prediction accuracy of EM algorithm is the highest. It shows that EM algorithm is reasonable and feasible in incomplete measurement data processing.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU196.1

【参考文献】