Study On The Dynamic Deformation Data Analysis Model Based On EMD

High precision deformation monitoring is of great significance to the health maintenance of the structure. Satellite positioning technology is widely used in high precision monitoring of large structures, but it is subject to the influence of high frequency noise and low frequency multipath noise, and the deformation monitoring precision is hard to meet. In order to extract the real deformation from the nonlinear and non stable monitoring signals, this paper introduces the method of empirical mode decomposition of the modern signal processing technology. On the basis of the whole theory of EMD, a lot of reasearch work is done about the problems of end effect and mode mixing in EMD algorithm. The improvement measures to restrain the end effect and the mode mixing problem are put forward. In the research of the real deformation data denoise and prediction algorithm and the development of deformation system, some research work has been done.The main research content are as follows:(1) Improvement of EMD endpoint effect and mode mixing problemThe end effect of empirical mode decomposition(EMD) was discussed. On the basis of analyzing the ordinary boundary extending methods, to fully exert their advantages and maintain maximum signal intrinsic trend, a self-adaptive extension algorithm based on feature extraction was proposed. Waveform variation law was summarized based on feature extraction. Using the improved template matching algorithm to get the consistent waveform with the endpoint. Using the improved triangular matching algorithm to get the similar waveform with the endpoint. When the signal regularity is poorer, failed to select similar waveform, we use RBF neural network to extend. Through the simulation data, this method can be used to realize the smooth transition of the extended data and the original signal waveform. Aiming at the problem of mode mixing in EMD algorithm, Complete Ensemble empirical mode decomposition(CEEMD) is studied in this paper, which is newly variation of EMD. It can better spectral separation of the modes and keep the completeness of EMD method.(2) The research on denoise processing and prediction of time series dataIn order to eliminate the noise in the deformation sequence, the signal is decomposed into different scales by using CEEMD method. Aiming at the problem that the signal and noise distinguish criteria are not unique, the denoising algorithm based on the combination of CEEMD and auto correlation analysis is proposed to separate the signals and random signals. The algorithm is applied to the simulation experiment and GNSS deformation monitoring data, and the traditional wavelet denoising methods are compared and analyzed. Compared with the wavelet method, the algorithm has a better effect. In order to better analyze the deformation law and extract the deformation information, this paper studies the prediction of the deformation signal. A hybrid RBF algorithm with nearest neighbor and gradient descent is used to improve the structure of neural network, and the effect of the algorithm is tested by the measured signal after noise reduction. Compared with ordinary RBF, the efficiency is guaranteed and the accuracy of prediction is improved.(3) Dynamic deformation monitoring data processing and analysis systemAccording to the requirement of deformation monitoring data processing, design and implementation dynamic deformation monitoring data processing and analysis system. Based on ASP.Net and Matlab mixed programming technology, the system construction the core algorithm components of the time series data reduction and deformation prediction of observation data. Based on Arc GIS Java Script APIs technique, the spatial visualization expression of deformation monitoring data is realized, and the basic spatial query and analysis functions are provided. The experiments show that the system can achieve better processing, analysis and display of deformation data.