上海交通大学学报 ›› 2023, Vol. 57 ›› Issue (8): 963-971.doi: 10.16183/j.cnki.jsjtu.2022.089

所属专题: 《上海交通大学学报》2023年“船舶海洋与建筑工程”专题

• 船舶海洋与建筑工程 • 上一篇    下一篇

一种CNN-LSTM船舶运动极值预报模型

詹可1,2, 朱仁传1,2()   

  1. 1.上海交通大学 船舶海洋与建筑工程学院
    2.海洋工程国家重点实验室,上海 200240
  • 收稿日期:2022-03-31 修回日期:2022-05-27 接受日期:2022-07-27 出版日期:2023-08-28 发布日期:2023-08-31
  • 通讯作者: 朱仁传,教授,博士生导师,电话(Tel.):021-34204288;E-mail:renchuan@sjtu.edu.cn.
  • 作者简介:詹 可(1997-),博士生,从事船海工程水动力学研究.

A CNN-LSTM Ship Motion Extreme Value Prediction Model

ZHAN Ke1,2, ZHU Renchuan1,2()   

  1. 1. State Key Laboratory of Ocean Engineering
    2. School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
  • Received:2022-03-31 Revised:2022-05-27 Accepted:2022-07-27 Online:2023-08-28 Published:2023-08-31

摘要:

针对船舶摇荡运动的短期极值预报,提出一种融合运动频谱信息的滑动窗口方法提取特征数据,在此基础上构建卷积神经网络(CNN)与长短时记忆(LSTM)网络的组合预报模型,其中CNN模块针对输入数据的局部相关特征,LSTM模块针对数据的时间维度特征.对S175船进行运动仿真测试,结果表明,该模型对未来1~2个周期内的运动极值信息预报效果良好,各项评价指标均明显优于LSTM和门控循环单元(GRU)模型,具有重要的应用价值.

关键词: 局部极值, 短期预报, 卷积神经网络, 长短时记忆网络

Abstract:

Aimed at the short-term extreme value prediction of ship motion, a sliding window method based on motion spectrum information is proposed to extract feature data, based on which, a series prediction model of convolutional neural networks (CNN) and long short-term memory (LSTM) is built. The CNN module aims at the local correlation characteristics of the input data, and the LSTM module aims at the time dimension characteristics of the data. The simulation test results of S175 ship show that the model has a good prediction effect on the motion extremum information in the next 1 and 2 cycles, and the evaluation indexes are significantly better than those of LSTM and gate recurrent unit (GRU) models, which has an important application value.

Key words: local extremum, short-term prediction, convolutional neural network (CNN), long short-term memory (LSTM) networks

中图分类号: