基于CNN的波浪中船舶摇荡运动智能预报  

Intelligent Prediction for Oscillation Motion of Ships in Waves Based on Convolutional Neural Network

作　　者：林辰儒 华先亮 谢立新 王扬理 侯先瑞 LIN Chenru;HUA Xianliang;XIE Lixin;WANG Yangli;HOU Xianrui(Shanghai Ship Design and Research Institution,Shanghai 201303,China;College of Ocean Science and Engineering,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海船舶研究设计院,上海201303 [2]上海海事大学海洋科学与工程学院,上海201306

出　　处：《船舶工程》2024年第4期44-53,共10页Ship Engineering

基　　金：大型远洋船舶智能航行技术研究(CBG4N21-2-4)。

摘　　要：准确预报波浪中船舶的摇荡运动对保障船舶的航行安全和正常营运具有重要意义。研究应用卷积神经网络(CNN)对船舶在波浪中的垂荡-纵摇耦合运动进行预报研究。对船舶在不同规则波激励下的垂荡-纵摇耦合运动进行分析,对比CNN与长短期记忆(LSTM)神经网络的预报结果,验证CNN模型的预报性能。数值模拟得到船舶在白噪声谱和JONSWAP谱激励下的不规则垂荡-纵摇耦合运动响应,应用CNN模型对所构造的训练集进行学习,并对测试集进行预报。对比CNN与LSTM的预报结果,检验CNN在不规则波中船舶摇荡运动方面的预报性能。结果表明:CNN和LSTM神经网络具有同级预报精度,可对船舶在波浪中的垂荡-纵摇耦合运动进行准确预报。It is very important to accurately predict the ship's oscillating motion in waves to ensure the ship's navigation safety and normal operation.A convolutional neural network(CNN)is applied to predict the heave-pitch coupled motion of ships in waves.The ship heave-pitch coupled motion under different regular wave excitations are analyzed,and the prediction results of the CNN and long short term memory(LSTM)neural networks are compared to verify the prediction performance of the CNN model.Secondly,the ship's irregular heave-pitch coupled motion response under the excitation of the white noise spectrum and the JONSWAP spectrum is obtained by numerical simulation.The CNN model is used to learn the constructed training set,and the test set is predicted.The prediction results of CNN and LSTM are compared to verify the prediction performance of CNN in a ship's rocking motion in irregular waves.The results show that CNN and LSTM neural networks have the same level of prediction accuracy,which can accurately predict the ship's heave and pitch coupling motion in waves.

关 键 词：船舶 摇荡运动 运动预报 卷积神经网络 智能航行 

分 类 号：U661.32[交通运输工程—船舶及航道工程]