Study Of Dam Safety Monitoring Models Based On Neural Network

Dam safety monitoring is an important means to master the working state of dams, and also the scientific basis to judge the safety degree of the dams. Analysis and applications of theory and methods for dam safety monitoring, which play an important role in guaranteeing the safety operation of dams, have been made considerable progress. However, there are still some problems and deficiencies. In this paper, the existing models of dam safety monitoring and dam safety evaluation are analyzed. And aiming to the defects in traditional methods, the BP neural network model is optimized combining to theories and methods in some other fields, and is applied to dam safety monitoring and dam safety evaluation.The main contents are as follows:(1)The basic features, network principles and implementation steps of BP neural network are introduced, and in order to improve the defects of BP neural network, some measures are put forward.(2)The building process of statistic model and the theory of multivariate linear regression analysis are introduced. The BP neural network model of dam safety monitoring is built and based on it, a BP neural network merging model is put forward creatively. Case study shows that the RMSE of the statistical model and BP neural network model are ±0.494mm and ±0.414mm, when the RMSE of BP neural network merging model is ±0.325mm. The model accuracy of merging model are better than the statistic model and BP neural network model.(3)In order to improve the defects in BP neural network such as poor stability, influence of initial values and local minimum, the genetic neural network model is constituted combining genetic algorithm and BP neural network. Then based on it, the genetic neural network merging model of dam safety monitoring is put forward. Case study shows that the accuracy of genetic neural network model is improved compared with the BP neural network, and the genetic neural network merging model is best.(4) Based on the traditional dam safety evaluation models, the concept and calculation formula of safety degree value are put out and the genetic neural network model of dam safety evaluation is proposed. Case study shows that the model has a higher precision with a small difference between the fitting results and the actual calculation results of the dam overall safety degree value, the biggest absolute value of the difference is only 0.0403. And the dam safety evaluation can be carried out directly with original dam observation data by this method.