Research On Filtering Algorithm In Deformation Monitoring With Prior Constrained Information

In the data processing of the deformation monitoring, constraint information can be properly used under certain objective conditions, which can apparently simplify the mathematical model and improve the accuracy of the estimated deformation parameters as well as control the filter divergence. But as the state constraints changes the probability structure of state parameters in the kinematic positioning, the difficulty for data analysis and filtering solutions will therefore increase. In the actual data processing, the commonly used method is to remove some state parameters by the state constraint equations, and then the left state parameters are estimated in a common filter processing. However, this tends to increase the complication of the calculation for some nonlinear cases and cause the greater change of the original filter equations. As a result, it is inconvenient in the practical application. This thesis investigates the acquisition of priori constraint information, conversion and the establishment of constraint filter model through a case study with an emphasis on how to build the mathematical model of deformation observation filter by using the geological, mechanical information of landslides and ground deformation information. Besides, based on the analysis of the current status and existing problems of the filtering algorithm, some novel algorithms have been proposed in terms of different constraint information. The new algorithms can increase the validation of the solutions when with abnormal or insufficient measurement information. The main contributions of the thesis are given as follows:1. Discuss the acquisition of the priori constraint information, establishment of model and summarize the forms of various constraints and build the corresponding filter model through a case study, focusing on data processing with the consideration of the landslide with geological and mechanical information and result analysis on the corresponding model and algorithms. 2. For parameters with adjustment model of inequality constraint information, the general form of least square solutions is given, which facilitates the implementation of constraint filtering algorithm in deformation observations.3. For some priori constraint information with unemployed unknown parameters, an algorithm is provided to convert it into state equation containing unknown parameters. Then the fitting method of systematic errors with moving window is adopted.4. A filtering algorithm with a proper use of geometry information and physical information is given. Firstly, some unknown physics information is thought as unknown systematic errors, which are then estimated in the data processing. Secondly, the priori constraint information is used to control the influence of the geometric observation abnormalities on deformation parameters.5. A filtering algorithm with the capability of resistance to blunders is proposed, in which the state parameters are constrained by constraint information. The filter solution without constraint firstly provides initial state estimates and then the state estimates are updated using comprehensively constraint information.In the data processing of deformation observation, the displacement obtained by the forecast of physical model of deformation body is used as a priori constraint. By handling the geometric measurements by continuous filter method, and adjusting the physical model information and relative weight between the physical model information and geometric measurement information, and modifying the parameters of physical model of the deformation body, the impact of the errors can be minimized and therefore optimal deformation parameter estimates can be obtained. The thesis innovatively solves the use and calculation issue of a priori constraint information and enables the application in the data processing of deformation observation. Meanwhile, the filtering theory is extended to enable the processing for the case with the prior constraint information and the data processing method of kinematic filtering is further improved and developed.