The Application Of Improved Particle Swarm Optimization BP Neural Network In Dam Deformation Prediction

As dam construction engineering according to its construction, hydrology,geography, address conditions very own architectural characteristics, we can know the dam project is a very complex project, and susceptible to external environmental factors, at the same time there is a certain social security risks when the dam project makes in the service. So,for such a hydraulic dam building, the security and safety monitoring more and more attentions. Currently, the dam safety monitoring have a number of methods, mainly all the real-time monitoring of the dam, through a lot of deformation monitoring data, to analyze the situation and determined the safe operation of the dam. Research Methods in which the dam deformation forecasting is usually only a single prediction model, and after a lot of practice studies have shown that a single dam deformation prediction model has a lot of flaws and shortcomings,and has a big difference between the predicted results, so the reliability and stability is not so high.Based on the BP neural network model in dam deformation were divergent research, and there are many shortcomings and deficiencies at the dam deformation prediction by experimental studies demonstrate a large number of single BP neural network model. For which these defects and deficiencies in the forecasting process,and proposes a new way of thinking which to use a strong characteristic of intelligent optimization particle swarm optimization in the single BP neural network model. Its design ideas is to use particle swarm algorithm to optimize the weights and thresholds of BP neural network, which not only can optimize the structure of the neural network,can also improve the efficiency of the neural network. Then, from the basic theory of particle swarm starting to improve the performance of particle swarm optimization.Particle swarm algorithm proposed by the previous presence of premature convergence in the process of optimization problems, so we introduce an adaptivemutation in classical particle swarm algorithm to avoid particle premature convergence thereby falling into the local minima. And add a random in the initial position and velocity of a particle, this can increase the particles in the global search,global search capability can be improved particles. After the above operations can be improved greatly improved particle swarm optimization to optimize the efficiency of the algorithm. Finally, this improved particle swarm optimization BP neural network model to engineering practice, through experimental studies show that the model played a good effect of the dam deformation prediction, and the model can be applied in dam project monitoring data analysis and forecasting.