The Research In Improvement Of The Grey Model Based On Kalman Filtering And Application Of Deformation Prediction

Deformation monitoring is an effective mean to accurately analysis and forecastbuildings and constructions’ real deformation, whereas in modern human society, for the rapiddevelopment of our economy and society, high-rise buildings and constructions will be more,with this it develops a series of problems such deformation, and which is also the hot topic inengineering, how to accurately predict buildings’ deformation tendency is a complicated andsophisticated science difficulty, no doubt, if we can exactly get buildings’ deformationtendency, it will be useful, not only can we decline the disaster loss caused by deformation,but also we can take some measures to save many lives, in a word it’s meaning is deepest. Thethesis is based on the predecessors’ research of deformation prediction model, by exploration,I use the adaptation-Kalman filtering theory, grey theory, robust estimation theory, data fusion,variance component estimation, and other theories to construct the dynamic adaption Kalmanfiltering-Grey integrated correction forecasting model, at last, the writer use an engineeringexample to analysis the feasibility of the model, and the thesis is mainly to display the followcontent:(1) The technology of the Kalman filtering can weaken the influence of measurementnoise to observed value, and it can get the best estimated value from measurement data withnoise, it makes the quality of the data to construct GM(1,1) model be improved.(2) The thesis in the process to construct GM(1,1), which is using robust estimation toestimate parameters and replace least-squares estimation, and so the constructed GM(1,1)model has stronger robustness to traditional GM(1,1) model.(3) By using the residual error that from simulation data and constructed model data toconstruct GM(1,1) model of residual error, and it makes the quality of prediction data better.(4) The thesis use different mount of measurement data to construct several GM(1,1)models and use the technology of data fusion to get the best estimated data.(5) The thesis uses the residual error between prediction data and measurement data toalter dynamic nosie variance of Kalman filtering, and in some extent, it can improve theaccuracy of random-model, then it will improve the filtering data’ quality.(6) Based on the dynamic adaptive Kalman filter-GM comprehensive correction model, the thesis selects an engineering case to testify the model’ feasibility.