论文部分内容阅读
挑流冲刷的深度直接关系到大坝的安全,是泄洪消能工设计的首要依据。为此以200多组原型观测资料为依据,建立了旨在预测冲坑深度的改进向后传播(BP)神经网络模型和广义回归神经网络(GRNN)模型,并对影响BP模型精度的网络拓扑结构、数据处理方式以及网络学习算法进行了分析。利用这两种模型对10个工程的冲坑深度进行了预报,并与传统预报公式的计算结果作了比较。结果表明:这两种模型都能比较准确地对冲刷进行预报,并各自在一定范围内占优;如果将二者联合使用,则预测结果明显优于传统公式。
The depth of pick-flow erosion is directly related to the safety of the dam, which is the primary basis for flood-dissipator design. Based on more than 200 sets of prototype observation data, an improved back propagation (BP) neural network model and a generalized regression neural network (GRNN) model are established to predict the depth of scouring pit, and the network topology Structure, data processing methods and network learning algorithms were analyzed. The pit depth of 10 projects is predicted by using these two models and compared with the results of traditional forecasting formula. The results show that both models can predict erosion more accurately and each occupy a certain range. If the two are used together, the prediction result is obviously better than the traditional one.