均匀沙床面动床阻力计算的人工神经网络模型

针对泥沙运动数据含有噪音且样本数据较多的特点，提出采用人工神经网络(ANN)BP模型批学习的训练方法可有效地缩短计算时间、提高训练精度。建立了由能坡、无因次单宽流量和无因次泥沙粒径等3个参数预测水深的结构为3-33-1的冲积河槽动床阻力BP模型。预测结果与实测资料比较表明这个人工神经网络模型的预测精度较高，同时这个模型与21个动床阻力公式的比较表明人工神经网络模型要比传统的回归模型精度高。

ANN model for flow resistance of movable bed formed by uniform sediment

Sediment transport in alluvial channels is a problem of nonlinearity, and artificial neural net (ANN) have many merits in dealing with nonlinearity. For abundant sampled data which contain noise from measurements and sediment transport itself, batch-mode training can effectively reduce the computing time and improve the training precision. The A3-33-1 ANN model is established for the prediction of flow depth according to energy slope, dimensionless unit discharge, and dimensionless sediment diameter. Compared with the measured data, the results show that the ANN model has a high precision. Also, compared with other equations based on regression, the ANN model is better than traditional regression model.