Research On Optimization Of Dam Safety Monitoring Statistic Models Based On Genetic Algorithm

The factors which affect the dam security are multiple and complex,so the urgent problem in the research of the dam security monitoring model is how to establish a reasonable dam security monitoring model to reflect the changing rules between loading effect set and loading set.Currently,the deformation monitor mainly adopts the statistical model,deterministic model and hybrid model.The statistical model is more applied due to its simple calculation.However,it is usually difficult for the traditional statistical model to effectively and accurately describe the relationship between each effect factor and the deformation because of the terrible operation environment of the dam and excessive monitoring effect factors.Therefore,in order to establish accurate and reasonable concrete dam deformation model,this paper makes deeper research on the statistical model of dam using stepwise regression analysis,genetic algorithm,fuzzy logic inference and some other methods,the main contents are as follows:Firstly,this paper comparatively analyzes the features for establishing the statistical model of concrete dam deformation using the multiple linear regression and stepwise regression method.For the stability and the approximation accuracy issues existing in the multiple linear regression method,this paper uses the stepwise regression analysis to optimize the regression factors in normal statistical model and establishes the statistical model for dam displacement monitoring using the loading set which has bigger influence on the dam deformation.The stepwise regression analysis method solves the multiple linear correlation problems of the factors of loading set and it can obtain higher approximation accuracy compared with multiple linear regression methods.Secondly,for the poor fitting problem existing in the statistical model which is established on the basis of the stepwise regression analysis,this paper makes a global search for the model regression coefficient using genetic algorithm in order to obtain the global optimal solution.In order to improve the approximation accuracy of the genetic algorithm,this paper designs a kind of adaptive genetic algorithm which can automatically adjust the crossover probability and the mutation probability according to the difference between the average population adaptation degree and the optimal adaptation degree in the process of the genetic algorithm search,avoid the genetic algorithm process falling into the local minimum and improve the approximation accuracy of the dam model regression coefficient.Thirdly,for the disadvantage of the excessively slow convergence rate for genetic algorithm during the evolutionary process,this paper proposes a kind of improved algorithm of the population diversity based on the fuzzy logic inference,and this fuzzy logic inference system adjusts the genetic algorithm operation according to the population variance and the entropy dynamically in order to keep good diversity of the population.For the selection methods on the genetically operational parent individuals,this paper proposes a method combining the optimal storage strategy and uniform order which can guarantee the global convergence,improve the searching efficiency and solve the problem of the excessively slow convergence rate for genetic algorithm during the evolutionary process.Finally,it comparatively analyzes the statistical performance index of all kinds of models.Adopting the genetic algorithm optimal model under the same effect factors is better than the stepwise regression model both on fitting precision and predicted accuracy.