A Study on Roughness Coefficient Using BP Neural Network

Since 1999, Xiaolangdi reservoir plays an important role in flood control, irrigation and  repair and maintenance of the healthy life of Yellow River. At the same time, process which the water and sediment flow into the downstream has been changed by the regulation of reservoir and trigger a number of new phenomenon. The abnormal phenomenon that a flood peak increased in August 2004 , July 2005, August 2006, August 2007 along the lower Yellow River occurred after the density current is poured. The fundamental reason for this phenomenon is the decrease of integrated roughness coefficient. Comprehensive roughness coefficient is an  important parameter for the river flow dynamics and mathematical model,whose correct or not directly influence the accuracy of the model. After analyzing the factors influencing roughness, a BP neural network model is built to calculate the roughness. Median grain size of bed load, sediment concentration, median grain size of suspended load, Froude number is the input of the model, the roughness coefficient is the output of the model. Through the verification of  the roughness coefficient in the course of  the "04.8", "05.7", "06.8", "07.8", the results show that the neural network model can calculate roughness coefficient accurately.