The Safety Monitoring Model Research Of Earth-Rock Dam Based On Genetic-Regression And BP Network

With the development of exploitation and utilization of water resource, the safety of dam becomes increasingly serious. There are lots of dams, especially the earth-rock dam, happened cracked before. The earth-rock dam is the main type dam in our country and the mostly proportion of diseased reservoir. Therefore, to dispose and analyze the measurement data in time for earth-rock dam and its characteristics, and to establish forecast model for quantitative analyzing based on qualitative analysis, is necessary to realize the real-time judge of dam safety situation to feedback and control the safety and running of dam, and is great significance to search after more analyzing method of monitored date and the fited actual monitoring model. At present, much research has been done on certain aspect (such as on distortion or seepage) of prototype monitoring data of earth-rock dam. But the analytical methods are limited in the traditional regression analytical method or single calculation method such as genetic algorithm or neural network to build the model. Therefore, the text has studied the monitoring model of prototype monitoring data of earth-rock dam with several analytical methods. The main research contents and results acquired can be summarized as follows:(1) The general theory and process of prototype monitoring data are studied. The stepwise regression analytical method calculation modeling program is built to analyze the relationship between effect quality and influence quality of earth-rock dam.(2) The genetic-regression model is built. The influence fact volume of earth-rock dam effect quality is gotten by stepwise regression analysis. The genetic algorithm is used to randomly select a set of best combination of independent variable factors in the forecast factors volume, while the root mean square errors of fitting and forecast of multi-linear regression equation combined with the selected factors are least. The model which is applied on a calculation example is proved to be feasible and has rather good precision. (3) The BP neural network model is built. The model and structure of BP neural network and its correlation math expression deducing, the principal and design of BP network faced MATLAB are discussed in detail. The BP network model is built by adding a momentum term to the variations of each power value and threshold value during the process of study training to improve BP neural network. The calculation program of BP neural network model is built with MATLAB and applied to engineering example. The calculation results show that the fitting and forecast effects of the model are better thfan the previous two models.(4) Engineering application research. Three types of forecasting model are built and applied on an engineering which results are analyzed comparatively by drawing and listing. The advantages, disadvantages and applicable conditions of each method are discussed to supply reference and more choices while analyzing data of real engineering.