摘要
结合船舶操纵运动中横摇运动的特点,建立了基于径向基神经网络的四自由度船舶操纵运动预报模型。通过Z形试验和回转试验的部分仿真数据充分训练该神经网络,然后使用训练好的网络预报10°/10°Z形操纵运动,并将预报结果与仿真值进行对比,验证了所提出预报方法的有效性。同时基于该网络预报20°/20°Z形操纵运动和35°回转运动,证明该预报方法的具有良好泛化能力。
Considering rolling motion characteristics of ship manoeuverability, a prediction model of 4 degrees of freedom for ship manoeuvring motion based on neural network with radial basis function is established. The neural network is trained with one percent of simulation data (5/5, 10/10, 15/15, 250/25, 300/30 zigzag tests and 15, 25 turning circle manoeuvre) . The 10/10 and 200/20 zigzag tests motion and 35 turning circle manoeuvring motion are predicted using the trained neural network. The predicted results are compared with those of the simulation tests to demonstrate the accuracy and the generalization ability of the proposed method.
出处
《中国造船》
EI
CSCD
北大核心
2013年第4期155-162,共8页
Shipbuilding of China
关键词
船舶操纵
运动预报
神经网络
四自由度
ship manoeuvring
motion prediction
neural network
4 degrees of freedom