The Application Of Real-coded Genetic Algorithms In Network Adjustment And Prediction Of Deformation Monitoring

Deformation monitoring is one of the most important safety diagnosis methods for monitoring body. With the improvement of modern building’s complexity and the enhancement of people’s security awareness, the work of deformation monitoring is more and more valued by people. The rapid development of science and technology makes high precision and high quality deformation monitoring to become real, and it also puts forward a higher request on data processing method of deformation monitoring. Using geodetic method to process deformation monitoring data mainly includes three aspects of the content: adjustment computation of deformation monitoring network, deformation analysis (stability analysis, physical explanation) and deformation prediction. Those three aspects include many nonlinear mathematical models, such as adjustment model, prediction model and physical model. However, in the actual work the data process method of deformation monitoring also use the linear method, it obviously has failed to meet the development of the deformation monitoring technology and the request it’s service object. Therefore, how to use nonlinear theory and methods to process deformation monitoring data become a hot area of research and this also is the major research content of this paper.(1) This thesis from the adjustment datum of deformation monitoring network detailed analyze the internal relations between the classics adjustment model and datum, and on the based, this thesis establishes the uniform adjustment model of deformation monitoring by added equation condition of datum, systemic summarized the computational formula and relevant parameter of the model, and deduced the transformation model between the different datum’s adjustment results. The actual engineering calculation proves the validity and practicability of the uniform model.(2) In allusion to the urgently needs of nonlinear method on data processing of deformation monitoring, this thesis deeply researched the mathematical theory and evolutionary mechanism of genetic algorithms which has global optimization characteristics, comprehensively analyzed the genetic operator parameters does how to influence the global convergence performance of genetic algorithms, and proposed dimensionality reduction techniques to solve the problems of COPs by genetic algorithms. According to the practical problems on network adjustment of deformation monitoring, this thesis improved the control parameters of conventional genetic algorithms, and availably enhanced the optimizing performance. (3) The improved genetic algorithms are for the first time used in adjustment computation, which realizes nonlinear adjustment of deformation monitoring network under different datum. This thesis detailedly expounded genetic algorithms combining lest square principle’s algorithm design and implementation, proved the availability of adjustment computation by improved genetic algorithms and the superiority compared with the linear adjustment.(4) This thesis detailedly expounded the modeling process and the method of level judgment of gray GM(1,1), and deeply analyzed the white background value (λ) does how to influence the modeling precision. According to the defect of traditional λ value, this thesis use the above improved genetic algorithms to optimize it, and proved the higher prediction accuracy and stronger adaptability of optimized GM(1,1).(5) Based on the above research, this thesis developed the software of GA_DMAP which can achieve the nonlinear parameter adjustment based on genetic algorithms of deformation monitoring data and predict the deformation trend by using Visual C#programming language.