摘要
对利用神经网络预报平底结构入水砰击压力的方法进行了探讨。首先利用仿真软件计算各种情况下平底结构入水所产生的砰击压力,以此形成训练神经网络的数据集。其次利用数据集对三层反馈式网络进行了训练,讨论了不同隐含层节点数对该非线性系统的拟合能力,并且对梯度下降法、动量修正法和基于优化的LM算法的有效性和精度进行了比较,最后得出了适合平底结构入水砰击系统的网络结构。
This paper discusses the issue about the prediction of the peak value of slamming pressure of a flat-bottom structure by means of Neural Network. At first we utilize the simulation software to calculate the slamming pressure of a flat-bottom structure entering water go as to form the data group for training the Neural Network. Then we discuss the different structure forms of the network with different node numbers, compare the precision of different study rules and thus get the reasonable network form fit for the flat-bottom structure entering water system.
出处
《海洋工程》
CSCD
北大核心
2005年第2期26-31,41,共7页
The Ocean Engineering
关键词
神经网络
平底结构
入水
砰击压力预报
NN
flat-bottom structure
entering water
slamming pressure prediction
