Application Of Neutral Network Based On Genetic Algorithm In Dam Displacement Forecast

Dam is the building which is a complex structure and endures enormous loads. As the body becomes huger and higher, the safety of the dam has taken more and more attentions among the hydraulic workers. Dam deformation forecasting is a crucial component part of the safety monitoring system and plays a significant role in the respect of the operations security. Security analysis and comprehensive assessment of the dam are based on the safety monitoring model. The traditional monitor systems generally adopt statistics model,assured model or mixed model. Without question, these classical models had shown their significance in respect of dam deformation forecasting in recent decades . However, the working condition of the dam is complex and each dam has its own character, there are still some unsure factors which can't be described by certain ration relation. So these traditional models can't make a complete description of the non-linear relationship of the variables. As a result, it will reduce the precision of forecast and matching. For the badly anti-jamming capability, these models may cause a non-ideal result as long when the data is few or the errors are large.Artificial neural network, as an artificial intelligence technique, develops rapidly in the recent fifty years, and attains remarkable breakthroughs. It is mainly applied in pattern identifying, image processing, automation and optimizing, forecast and intelligent information managing, communication and so on. Its virtues which contains organization, adapt capability, associational capability, fuzzy ratiocinative capability and self-educated capability are suitable to solve nonlinear problems. So in the recent few years, many hydraulic workers applied it in the filed of dam safety monitoring, and utilize the function approach capacity of artificial neural network to simulate the complicated relationship of the correlation quantity of the circumstances and the consequent variable in dam safety monitoring. And it achieves the function of forecasting the consequent variable according to the independent variable, and attain excellent effect.In this paper, BP neural network arithmetic is studied first. BP neural network has the characteristics of initialization randomicity,convergence slowly,easily convergence to local least value and so on. Combined with the results of the past research, Levenberg Marquardt arithmetic is applied to the improvement of classical BP neural network. What's more, the paper applies automatically Adaptive Genetic Algorithm to global optimize the weight value and threshold value of ANN. The model built above can efficiently improve the simulating effect and forecasting precision.In order to affirm the GA-LMBP model applied in this paper, we utilize the transverse displacement monitoring data of Jilin Fengman Dam. Apply MATLAB to simulate it, and then contrast the result with that attented by linear regression analysis model and classic BP neural network model. Consequently, we get a conclusion that the model applied in the paper is more steady and more accurate, and it is worth to adopt.