基于GRNN的坡面径流输沙能力模型的试验研究

坡面径流输沙能力是建立土壤侵蚀过程模型的重要水力学参数,研究定量计算坡面径流输沙能力的实用模型具有重要的理论和实践意义.通过室内模拟径流冲刷试验,计算不同坡度和流量条件下的裸地坡面径流输沙能力,利用平均影响值(MIV)方法对影响坡面径流输沙能力的因子进行分析,建立以干密度、能坡、进口流量、出口流量、水力半径、流速为输入,以坡面径流输沙能力为输出的广义回归神经网络(GRNN)模型,并应用Adaboost算法对模型进行优化.验证结果表明,所建模型能够用于对坡面径流输沙能力的模拟预测.与BP神经网络模型进行对比分析的结果表明:在试验训练样本条件下,广义回归神经网络(GRNN)模型的模拟预测结果优于BP神经网络模型；Adaboost算法能够有效减小广义回归神经网络(GRNN)模型的模拟预测误差.

Experimental study on model of sediment transport capacity of slope runoff based on GRNN

Sediment transport capacity of slope runoff is an important hydrodynamic parameter in establishing model of soil erosion process,it is of theoretical and practical significance to study the model calculating sediment transport capacity quantitatively.By means of indoor simulated runoff-scouring experiments,sediment transport capacity of slope runoff under different slop and flow conditions was calculated.The impact factors of sediment transport capacity of slope runoff were analyzed by using method of Mean Impact Value.Generalized Regression Neural Network(GRNN) model was established,in which input variables include dry bulk density,slope,Inlet flow,outlet flow,hydraulic radius and flow rate,output variable is sediment transport capacity of slope runoff.Additionally the model was optimized by Adaboost algorithm.The validation results showed that the GRNN model was feasible to predict sediment transport capacity of slope runoff.Under conditions of experimental training samples,GRNN model performed better than BP Neural network model,and Adaboost algorithm could effectively reduce error in the prediction of GRNN model.