船载运动监测决策系统是船舶智能化、运动响应监测与预报中所涉及的重要技术，目前现有方法中仍采用幅值响应算子等方法进行运动预报，为解决预报解算效率、提高运动预报精度，本文提出基于深度学习的船载运动监测决策系统。文中针对一艘特定纸浆船开展了系统搭建与测试，通过姿态传感器的布置安装及加速度处理算法的软硬件架构，结合基于模型试验的预训练预报模型，对实船运行中的加速度运动进行推算及预报。结果表明，该推算及预报方法在船舶停泊和航行2种状态下均有较好的表现：运动推算中均值绝对误差不超过0.005 m/s2，标准差绝对误差不超过0.1 m/s2；运动预报中均值绝对误差不超过0.01 m/s2，标准差绝对误差不超过0.05 m/s2。本文提出一种具有普适性的船载运动监测决策系统，并结合深度学习方法进一步提高了预报的效率及精度，为未来进一步拓展系统功能及运动监测预报提供了软硬件基础和研究路线。
Shipboard system for motion monitoring and decision-making is an important technology involved in ship intelligence and motion response monitoring & prediction. Currently available methods still apply amplitude response operators and other methods for motion prediction. In order to solve the efficiency of prediction solving and improve the accuracy, a deep learning based shipboard system for motion monitoring and decision-making is proposed. In the paper, the system construction and testing was carried out for a specific pulp ship. With the hardware and software architecture of the layout and installation of attitude sensors and the acceleration processing algorithms, the acceleration motion in the real ship was calculated and predicted, combining with the pre-training prediction model based on model tests. The results showed that the calculation and prediction method possessed good performance in both mooring and sailing states: as for the calculation, the absolute error of the mean value did not exceed 0.005 m/s2 and the absolute error of the standard deviation did not exceed 0.1 m/s2; as for the prediction, the absolute error of the mean value did not exceed 0.01 m/s2 and the absolute error of the standard deviation did not exceed 0.05 m/s2. The article proposes a universal shipboard motion monitoring and decision-making system, combining the deep learning method to further improve the efficiency and accuracy of prediction, which provides the hardware and software foundation and research route for further expanding the system function and motion monitoring and prediction in the future.
2025,47(1): 177-184 收稿日期：2024-3-6
DOI：10.3404/j.issn.1672-7649.2025.01.031
分类号：U661
基金项目：海南省自然科学基金青年项目（521QN276）
作者简介：潘文寅(1998-),男,硕士研究生,研究方向为基于深度学习的船舶及海洋平台运动预报
参考文献：
[1] 贾锐, 曹凯, 徐秀龙, 等. 浅析智能船舶系统 [J]. 船舶标准化与质量, 2016, 36–40.
JIA R, CAO K, XU X L, et al. Analysis of intelligent ship system [J]. Shipbuilding Standardization & Quality, 2016, 36–40.
[2] ABB. OCTOPUS-Onboard 8 User Manual [M]. 2017.
[3] B. V. M F. MO4 TECHNICAL NOTE [M]. 2018.
[4] KAPLAN P. A preliminary study of prediction techniques for aircraft carrier motions at sea[J]. Journal of Hydronautics, 1969, 3(3): 121-31.
[5] TRIANTAFYLLOU M, BODSON M, ATHANS M. Real time estimation of ship motions using Kalman filtering techniques [J]. IEEE Journal of Oceanic Engineering, 2003, 8(1): 9–20.
[6] TRIANTAFYLLOU M S, ATHANS M. Real time estimation of the heaving and pitching motions of a ship, using a kalman filter [C]// OCEANS 81, 1981.
[7] DUAN W, HUANG L, HAN Y, et al. A hybrid AR-EMD-SVR model for the short-term prediction of nonlinear and non-stationary ship motion [J]. Journal of Zhejiang University-SCIENCE A, 2015, 16(7): 562–576.
[8] DUAN W, HUANG L, HAN Y, et al. A hybrid EMD-AR model for nonlinear and non-stationary wave forecasting [J]. Journal of Zhejiang University-SCIENCE A, 2016, 17(2): 115–129.
[9] 蒋啊芳. 基于神经网络的地震数据时间序列分析 [D]. 合肥: 安徽大学, 2011.
[10] 杨小兵, 徐勇, 彭磊, 等. 时间序列分析——广义回归神经网络组合模型在痢疾发病率预测中的应用探讨 [J]. 中国初级卫生保健, 2008, 22(8): 64–66.
YANG X B, XU Y, PENG L, et al. Discussion on application of ARIMA-GRNN combination model in shigella epidemic situation forecast [J]. Chinese Primary Health Care, 2008, 22(8): 64–66.
[11] GUO X, ZHANG X, TIAN X, et al. Predicting heave and surge motions of a semi-submersible with neural networks [J]. Applied Ocean Research, 2021, 112: 102708.
[12] GUO X, ZHANG X, TIAN X, et al. Probabilistic prediction of the heave motions of a semi-submersible by a deep learning problem model [J]. arXiv preprint arXiv: 211100873, 2021.
[13] SUN Q, TANG Z, GAO J, et al. Short-term ship motion attitude prediction based on LSTM and GPR [J]. Applied Ocean Research, 2022, 118: 102927.
[14] ZHANG T, ZHENG X, LIU M. Multiscale attention-based LSTM for ship motion prediction [J]. Ocean Engineering, 2021, 230: 109066.
