摘要
将人工神经网络技术应用于计算Morison方程中的水动力系数Cd、Cm,构造具有一层隐含层的BP神经网络,然后在BP网络中运用附加动量法和自适应学习速率进行改造,使得建立的网络模型的收敛性大为改善,减少了训练次数和训练时间。结果表明,计算结果可靠,可用于计算不同雷诺数Re、KC数以及粗糙度数k下Morison方程的水动力系数,从而使得在利用Morison方程计算小尺度结构物的受力更接近实际。
Morison equation＇s hydrodynamic coefficients are calculated by applying BP neural network which has one hidden layer. Additional momentum method and auto-adjustive learning rate to modify the network are used. Thus the convergence property of the network model is greatly improved,the times and the time of training are reduced. Results indicate this method is reliable and it can calculate Morison equation＇s hydrodynamic coefficients according to different Re KC and roughness number K. So the force which acts on the small structures calculated by Morison＇s equation accords with the fact much more.
出处
《中国海洋平台》
2005年第6期18-23,共6页
China offshore Platform
