基于概率神经网络的河床床面形态预测模型

Predicting Model of River Bed Form Based on Probabilistic Neural Network

判定冲积河流床面形态的经验与图形分析方法较多,但一般只使用水流和泥沙参数中的2个参数,其判别效果不甚理想。为克服传统分析方法的局限性,在分析冲积河流床面形态影响因子的基础上,建立了非线性映射能力强的概率神经网络床面形态预测模型。实例分析表明,该模型能反映冲积河流复杂因素及其变化对床面形态的影响,较好地实现输入与输出的非线性映射,其预测效果较好,为床面形态的预测提供了一种新的方法。

Although there were lots of empiric and graphic methods about bed form analysis of alluvial river, only two parameters of water current and sediment were used generally, and the judgement effect of the method was not ideal. In order to overcome the localization of traditional methods, a PNN predicting model describing the bed form with strongly nonlinear mapping capability is established based on analysis of influencing factors of the bed form. An illustration shows that this model can reflect the influence of complex factors and their changes on the bed form of allurial rivers. The model is not only stronger in nonlinear mapping relationship between the input and the output, but also has predicted better results, and thus it provides a new predicting method for describing a bed form.