Research On The Combination Model Of Deformation Monitoring Based On Grey Model And Time Series Analysis

With the development of science and technology and growth of economic strength, moreand more large buildings appear constantly. These buildings inevitably produce deformationin the effect of various factors. Therefore, to ensure the safety of the engineering building wemust carry on the deformation monitoring, and analysis to identify the condition of thedeformation. At present, there are many theoretical methods of the analysis and prediction ofthe deformation, of which various have their own advantages and limitations. In order to bringinto full play the advantage of all kinds of forecasting methods and reduce the impact oflimitations, combination forecast analysis method arises.Grey model and time series analysis model are single forecasting model used in theanalysis of deformation. The application of the GM (1,1) which applies to deal with the smallsample data is simple, but the accuracy of the prediction is not high; while the modelingprocess of time series analysis model which apply to analyze large sample data is complicated.If combine the two, play out their respective advantages will achieve the aim of improving theaccuracy of the prediction model. The characteristics of the grey model and the time seriesanalysis model are analyzed, and the linear and nonlinear combination modeling is proposed,and the prediction precision and applicability of these two kinds of combination modeling arecompared by practical analysis in the dissertation. The main research contents of thedissertation include:1、The GM (1,1) and the time series analysis model of the basic principle are introduced,both modeling process of the forecast are discussed in detail. The respective advantages anddefects of these two kinds of single model are analyzed emphatically which is the theoreticalbasis of establishing combination model.2、Non-linear and linear gray-timing combination model are set up.It revealed that theprediction accuracy of the two combination models is higher than a single prediction modelby calculation and analysis of some examples. In the meantime the prediction accuracy ofnonlinear combination forecast model is higher than that of linear combination model.3、Improvements have been done on the two kinds of combination model. The results showthat the prediction accuracy of the improved linear combination model is better than thenon-linear model in the short-term forecast; while in the long-term forecast the predictionaccuracy is equal between the improved linear combination model and the nonlinear model.When the nonlinear combination model being improving, we found that if one of the modelsis improved, the other models should be considered whether it is still acceptable, so the improvement of the models is greatly limited. Conversely, the former problem is notconsidered in the linear combination model, so the advantages of each single model can befully exerted. When the number of single model is huge, the advantages of linear combinationmodel are especially obvious.The style of combination of the grey model and the time series analysis model have beenperfected in this paper, the two kinds of establishment combination model are obtainedsatisfactory prediction results and the trains of thought have been widened for the furtherimprovement of the gray-timing combination model.