| In our country there are a great many large ,medium and minitype reservoirs and dams have safety problems of different degree, especially do the earth-rockfill dam. Therefore, it is very important that using even mature and suitable theory and method analyzing prototype observed data and evaluating the condition, according to prototype observed data of earth-rockfill dam.In this thesis by means of regression analysis, genetic algorithms, artificial neural network,grey theory and multi-pole fuzzy pattern recognition theory, together with the characteristics of earth-rockfill, it is researched the observed behavior of earth-rockfill dam. The main contents are as follows:1. The genetic analysis for error of the prototype observed data of earth-rockfill dam is carried out and recognition methods of error is introduced. The reason of hysteresis quality and qualitative analysis of piezometric level is analyzed. The statistic model of piezometric level is established through genetic analysis for seepage of earth-rockfill dam.The influence factor of settlement of earth-rockfill dam is analyzed. The selection of settlement model is discussed,according to the characteristics of earth-rockfill dam and the features of various functions. Combining the measured seepage data of LuHun dam, qualitative analysis is carried out.2. The regression analysis method is introduced,such as simple regression analysis, curve regression which can be changed into simple regression analysis, multi-variate regression analysis and stepwise regression analysis. In the light of the problem of conventional regression analysis method based on least-squares method being unable to resist gross error interference ,we adopt M-estimator in common use in robust regression, which can conquer the influence of outliers in observed data effectively. The theory and solution procedure of M-estimator is summaried.At last we proof the above theory and method by applying it to observed data analysis for piezometric level of earth-rockfill dam.3. Genetic algorithms is introduced and improved ,which in considered of the physical problem of earth-rockfill dam.In light of the search capability of genetic algorithms is not powerful,we introduce simulated annealing algorithms,and improve it in order to improve it's search efficiency. Combining the settlement characteristics of earth-rockfill dam, the model of genetic algorithms and genetic simulated annealing algorithms is established. Hydraulic engineering has a lot of influence factor,which is a complex and high nonliner problem.The BP network model of artificial neural network is introduced and the BP network model of seepage and settlement is established.A lot of analysis and research indicate the aging deformation of earth-rockfill dam has monotonicity. Point this feature, we indraft gray theory, create the now information model of the GM(1,1), by using its characteristics that it canreflect the engineer's dynamic change .Finally we analyze and proof the above theory and method by applying it to observed data.4. The observed behavior of earth-rockfill dam is a complex multi- hierarchy and multi-object evaluation problem.In light of the deficiency of traditional evalution method unable to reflect integral behavior,we present and profound discuss the research body system of observed behavior of earth-rockfill dam which include the location of integral evaluation index,the measuring method of integral evaluation index and the path of integral evaluation. The multi-pole fuzzy pattern recognition theory is indrafted based on the characteristic of earth-rockfill dam evalution ,the model of multi-pole fuzzy pattern recognition theory for earth-rockfill dam is built and we apply it to example which analysis show that it is reasonable and feasible.5. Applying observed data of Luhun dam and the research of this thesis?together with database programming of Visual Basic and the development technique of software, we develop safety monitoring analysis and evaluation system of Luhun dam. |
PDF Full Text Request