Study On The Combination Models In Data Processing Of Deformation Monitoring

There are more and more high-rise buildings in China, as the development of economy and the progress of the urbanization. It is neccessary to monitor these buildings in the process of building and usage, in order to ensure the usage and the security of them.The goal of the deformation monitoring is to make use of the accumulated deformation data, analyze and assure the deformation characteristics and the rules of the deformable body, then makes the forecasting to the deformation tendency. Nowadays, there are many fitting and forecasting models of the monitor data, for example, Regression Model, Timing Series Model, Grey Model, Kalman Smoothing Model, Artificial Neural Networks Model and Wavelet Analysis and so on. But the certain limitations exist in these sole models, in reality, all the factors of the influence distortion aren't containing in the sole models.Since Bates J.M.and Granger C.W.J. firstly stated the combination forecasting methods, the research of which has maken a big progress. Combination Model is to suitablely combine the different forecasting methods and make good use of the massages of each sole forecasting methods, in result to enhance the availability of fitting and forecasting datas as far as possible. The establishments of the combination models are correlated with the final goals, meanwhile the useful combination models are different according to the differnt forecasting goals.The paper takes theactual subsidence datas of SOHU Union South Building as a foudation and takes the multiple regression analysis model, the timeing series analysis model and the GM(1,1) grey theory model as sub-models of combination model. After mass data calculations, this paper analyzes the superiorities of the combination models constracting to the sole models in fitting and forecasting, discusses the different characteristics of the different modelling plans of the combination model in the fitting of the historical datas and the forecasting of the unknown datas, and concludes the partial characteristics and its serviceability of every kinds of the combination models.