小波分析与时间序列组合模型在变形监测分析预测中的应用研究

发布时间：2018-05-15 13:42

本文选题：变形监测 + 小波分析　；参考：《长安大学》2014年硕士论文

【摘要】：随着现代科技的快速进步和国民经济的迅猛发展，现代各种工程建设的进程与速度也大大加快，而且现在我们对工程建筑物的建设规模、精度等有了更高的要求，这样为了保证工程建设的安全运行，对于各类工程的变形监测工作就显得尤为重要，尤其是对于变形监测数据的分析处理，更是重中之重。目前，对于变形监测数据的处理主要集中在分析变形原因和预报未来变形两个方面，我们在有限的观测数据的情况下想要预测未来的变形情况，是有很大难度的，现在，一般是选择有效的数学模型，根据监测数据的时序特点进行预报。现在常用的变形监测数据处理模型主要有：回归分析模型，时间序列分析模型，灰色理论模型，人工神经网络模型，，卡尔曼滤波模型，小波分析模型等。其中各个模型都有自己的优点和不足，而变形监测又是多因素的形变因子的集合，有时候在变形数据中包含有多种因子，这样处理数据就需要多个学科的交叉融合，所需要的处理模型也是多种多样，单一模型的处理精度就可能会达不到要求。因此，我们现在一般常用组合模型的方式解决这个问题，组合模型就是利用每个模型的优点，有机结合使其能够更加有效处理各种变形监测数据，提高分析预测的精度。本文在查阅大量文献资料以及各种工程实例基础上，提出了利用小波分析和时间序列分析组合的方法来进行各种变形监测数据的分析与预报。时间序列分析模型是一种动态模型，对于各类变形监测数据有着很好的兼容性，但是在处理非平稳的时间序列数据时，存在着差分化剔除趋势导致删除有效数据而造成预测精度降低的问题。小波分析模型中的小波变换则是一种能够有效地从时序数据中提取误差的方法，小波变换通过对监测数据的分解和重构，能够很好地反映出监测数据中的变形趋势及特征，从而分离误差。基于此，本文用两种模型相结合，有效地解决了时间序列分析中的剔除趋势问题，这个组合模型在实际的变形监测数据处理中有着良好的实用价值。
[Abstract]:With the rapid progress of modern science and technology and the rapid development of the national economy, the process and speed of modern engineering construction has been greatly accelerated, and now we have higher requirements for the construction scale and precision of engineering buildings. In order to ensure the safe operation of engineering construction, it is particularly important for the deformation monitoring work of all kinds of projects, especially for the analysis and processing of deformation monitoring data, which is the most important. At present, the processing of deformation monitoring data mainly focuses on the analysis of deformation reasons and prediction of future deformation. It is very difficult for us to predict future deformation under the condition of limited observation data. Now, In general, an effective mathematical model is chosen to forecast according to the time series characteristics of monitoring data. The commonly used deformation monitoring data processing models are: regression analysis model, time series analysis model, grey theory model, artificial neural network model, Kalman filter model, wavelet analysis model and so on. Each model has its own advantages and disadvantages, and deformation monitoring is a set of multi-factor deformation factors. Sometimes there are many factors in the deformation data. The required processing models are also varied, and the processing accuracy of a single model may not meet the requirements. Therefore, we usually use the combination model to solve this problem. The combination model is to use the advantages of each model, organic combination can more effectively deal with all kinds of deformation monitoring data, improve the accuracy of analysis and prediction. On the basis of consulting a lot of documents and engineering examples, this paper puts forward a method of combining wavelet analysis and time series analysis to analyze and forecast all kinds of deformation monitoring data. Time series analysis model is a kind of dynamic model, which has good compatibility for all kinds of deformation monitoring data, but when dealing with non-stationary time series data, There exists the problem that the trend of difference differentiation and culling leads to the deletion of effective data and the decrease of prediction accuracy. Wavelet transform in wavelet analysis model is an effective method to extract errors from time series data. By decomposing and reconstructing monitoring data, wavelet transform can well reflect the trend and characteristics of deformation in monitoring data. Thus separating the error. Based on this, the problem of eliminating trend in time series analysis is effectively solved by combining the two models. The combined model has good practical value in actual deformation monitoring data processing.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU196.1

【参考文献】