Application Of Combined Model Of Grey Model And Time Series To Forecast The Subsidence Of High Building

With the acceleration of urbanization rate in recent years, there are more and more high-rise buildings in cities. In order to ensure the safety of the building, so as to grasp the buildings' deformation, it requires a long-term and precise deformation monitoring, and it is necessary to make an accurate analysis and judgment on the trend of the deformation body, to provide the basis for the construction of the building at different stages.In this thesis, it's based on the settlement monitoring project of a residential district. The thesis analyzed the deformation prediction of the building by using the measured data as the basic data:1. In view of the high building subsidence data which has the characteristics of deterministic trend of non-stationary time series, the non-equal interval grey GM(1, 1)-AR combination model is proposed to predict the deformation and settlement of high-rise buildings. The main idea of the thesis is using the non-equidistance grey GM(1, 1) model to fit the subsidence of deterministic trends and using AR model to fit the uncertainty of random residual. Finally, the predicted results are summed up as the final prediction results.2. Selecting the original data and smoothing data for modeling data, and selecting different dimensions of different time interval data forecast comparison, the thesis analyzed the influence of the three aspects, such as noise processing, interpolation period and data dimension, on the prediction accuracy.3. This thesis analyzed the influence of interpolation cycle on forecasting precision for the non-equidistant sequence, choosing 7 days, 8 days, 9 days, 10 days, 11 days, 12 days as a time interval for spacing; choosing the 8 dimension, the 9 dimension and the 10 dimension series, it analyzed the influence of the selection of the sequence dimension on the prediction accuracy.4. The thesis used the MATLAB software to compile the gray GM(1, 1)-AR combination model program, which can be more flexible to change the parameters, and analysis the impact of these parameters on the accuracy of the model.