Forecast, Based On The Bam Neural Networks In The River Bed Cross-section Pattern Recognition Bridges Water Damage

In allusion to the shortcomings of the actual flood forecast methods of small and medium bridges, such as complicated field work and difficult to predict the bridges lack of data, from the reasoning formula, the flood forecast models directly from rainfall to waterline are set up for the nine types of riverbeds of small and medium bridges divided in this paper. To predict the waterline with these models, the BAM neural network is used for the pattern recognition of riverbed profile to choose the appropriate model. The parameters in the model can be gotten with the data of rainfall and waterline of the bridge. Then the function between rainfall and waterline of the bridge is conformed. So the upcoming waterline of the bridge could be predicted if the rainfalls are given.Because there are rainfall stations along the railways,this method can be used in the built railway bridges. The rainfalls could be deliveried to the computer in time and the waterline and the scour depth could be predicted before the flood peak arrives to gain the short but precious time for fighting the flood. To those bridges without peculiar rainfall stations, the professional rainfall prediction could be gotten from weather department. The prediction precision will be improved with the development of the weather forecast science and instrument.A example and its results are provided to demonstrate that predicting the waterline with the method proposed in this paper could make the prediction more accurate and could meet the needs of actual production. The workload of the field investigation is largely reduced and the flood forecasting work becomes more simple and effective. In particular, the method has more practical value in the bridges lack of hydrological data.