Research And Application Of Debris Flow Forecast Based On Artificial Neural Networks

China is the one of the countries which are faced to the world's most serious threat by debris flows, Debris, flow disaster prediction in the practice of disaster prevention is significant; it can be directly serve the national economy. Hundreds of people have been killed by the debris flow each year in average, a serious threat to the national economy and sustainable development of society. Debris flows and the complex relationship of many factors. Economic losses caused by the debris flow is increasing, therefore, debris flow forecast is an important issue for the research.Artificial neural network prediction model for disaster has been the object for many scholars, because the neural network has a better function of approximation, and more approximation ability for RBF Neural network than any other neural network function.Debris flow disaster prediction development process and the current study situation of forecast model are learned,The Debris flow disaster theories of prediction are outlined in this paper,A new model of debris flow trend is established for short-term forecast. It proposed the debris flow formations of internal and external factors innovatively, geological factors and rainfall factors are quantified and superimposed into account the method, using the theory of fuzzy mathematics, the rainfall factor is optimized to make the model more accurate.Detailed the artificial neural network principle,particularly the RBF neural network principle,the learning algorithm,the training process and the applied in the feasibility of using in Debris flow disaster prediction. combined with neural network variable selection method, the new model training process to effectively reduce the number of variables in the practical application, it can greatly reduce the workload of data collection and provide a good practical value. In this paper, the experiments of RBF neural network are maded by MATLAB 7.0, the results showed that the MIV-RBF is much better than the traditional network both in time-consuming and the accuracy.