BP神经网络预测河湾最大冲刷深度

BP Neural Network Model-Based Prediction of Maximal Scour-Depth at River Bends

影响河湾凹岸最大冲刷深度的因素众多,而且这些因素的关系是非线性的.实现河湾最大冲刷深度预测的实质是建立一个非线性映射.实现这种映射的传统途径是在室内试验的基础上,采用量纲分析和多元回归的方式建立经验公式.根据BP(前馈)神经网络模型能逼近任何闭区间的连续函数的性质,在室内试验的基础上,尝试采用人工神经网络模型对河湾冲刷深度进行预测,并与经验公式的计算结果进行了比较.结果显示,BP神经网络能够更为准确地对河湾最大冲刷深度做出预测.

Many factors of non-linear relationships affect the maximal scour-depth at river bends. As a matter of fact, to forecast the depth is to establish mappings between the factors and depth. The traditional way is to find experiential formulas based on experiments and dimension analysis. The paper attempts to adopt back propagation （BP） neural network model to predict the scour-depth according to its characteristics that BP can approach any continuous function. Study results show that BP Model can do better than the experiential formulas.