The Research Of Debris Flow Prediction Based On The BP Neural Network Method

Our country is much the mountain kingdom, the rock masses break, the influence of geological structure, many mountain steep mountain, unstable rock structure, in addition to the forest area is not much, per capita green area only1.9mu, only1/5of the amount per for the world, and then to the earthquake, floods, landslides and other natural disasters had created "good conditions". In these disasters, the most influential, the destructive power of strong, to the people’s life and property of the debris flow is poses a great threat. Debris flow disaster to people’s life and property safety pose a serious threat, and restricts the development of local economy and society. Every year because the death tolls of the debris flow hazards are among the flood and drought death toll front row, economic losses from millions to billions of dollars. Therefore, the scale of the debris flows out of natural disasters before disaster and forecasting work has a very important theoretical and practical value.At present, for debris flow hazards forecast work is mainly use the statistics analysis of regression analysis and grey system of GM prediction model. But the data modeling regression analysis more; Grey forecasting model due to the introduction of grey derivative, make the established model prediction accuracy is not high, However, the BP neural network model has good generalization ability, in training can make the network continuous optimization, and the whole process of training, compared with the regression model for less, sample relative grey forecasting model, the prediction accuracy is higher. Through case study and using BP neural network model for prediction of debris flow disaster is a precision of prediction method.Although the BP neural network model fault rate is strong, and the prediction accuracy is higher, in theory we can be applied to the mud flow disaster happened to carry on the forecast, but network in training, momentum, adopted the parameters such as the previous factor value only by the experience or even try to get, so this paper use of debris flow disaster related data set up the network in training hidden layer node, the number of momentum, with convergence factors adopted the relation model between the number of training, and make the network of the parameters of the have value basis; In addition, the BP neural network model in the debris flow prediction spend more time, convergence at a slower speed. And for network convergence speed is every time influencing the right value adjustment of network and training the choice of transfer functions. Therefore, this article is the two aspects of neural network model to improve model, and its application to the prediction of debris flow, the results show that the improved will the artificial neural network model is applied to debris flow disaster, the prediction accuracy is still high, and network convergence speed is standard neural network model was obviously improved, and its forecast for debris flow theory and practical research has laid a good foundation.