Research And Application Of Combined Deformation Prediction Model Based On EMD

At present, there are many theoretical methods for the analysis and forecast of the deformation monitoring data. But each method has its scope and conditions, in order to reduce the affect brought by limitations of a single forecast method, and to take full advantage of each forecast method, combined methods for analysis and forecast have emerged.Time series analysis model and gray model are single forecast models commonly used in the analysis of deformation. Time series analysis model is suitable for handling smooth,normality, zero-mean time series,and gray model is suitable for handling monotonically increasing time series. However, the actual deformation monitoring data is non-smooth and non-linear, there are both monotonous increasing trend and short-term fluctuations in the cumulative change of monitoring objects.If we process by using a single forecast model,typically, the efficiency and accuracy can not be both good. If the cumulative change of the deformation can be decomposed into random fluctuations items and monotonically increasing trend items, and then forecast by time series analysis model and gray model separately,combine the two models, play their owe advantages, achieve the purposes of accurate forecast.And the Empirical Mode Decomposition(EMD) can achieve this decomposition. The article is based on this idea, put forward the method of time series analysis based on EMD-GM,analyze and compare though practical examples, the results show that this combined model has advantages over the single time series model and gray model, and verify its feasibility and applicability. The main contents of the article are as follows:1.The basic principles of time series analysis model and GM(1,1) gray model have been introduced, the scopes of these two models have been analyzed, the whole process of modeling and forecasting has been discussed in detail, and the theoretical foundation for the establishment of the combined model has been laid.2.Basic theories of Empirical Mode Decomposition(EMD) have been extensively studied,principles of signal decomposition through the EMD method have been analyzed, and the specific decomposition steps have been discussed in detail.3.The time series analysis based on EMD-GM model has been established, and the specific algorithm has been achieved by MATLAB language, through practical case studies, itcan be found that the accuracy of combined forecast model is higher than the single forecast model, the feasibility and applicability of the combined model have been verified.