实海域风浪流耦合环境下滑行艇呈现强非线性运动特性，传统操纵性数学模型预报误差较大，实船试验成本高且工况覆盖不足，制约复杂环境性能底数评估。现有灰箱模型存在环境干扰简化（忽略波浪记忆效应）、实海域验证缺乏两大局限。本文提出一种LSTM-DNN灰箱融合架构，基于MMG理论框架，通过LSTM分支学习历史数据捕捉波浪频域特性与累积效应，DNN分支补偿静水动力与动态流载荷耦合误差，针对LSTM分支贡献进行消融实验和敏感性分析。结果表明，以10 m滑行艇为例在2～3级时变海况验证下，相较传统理论模型和纯数据驱动模型，灰箱模型显著提升实海域操纵性在轨迹、首向角和回转直径上的预报精度，可为船艇航行性能数字化试验鉴定提供技术基础。
Accurate prediction of planing craft maneuverability in coupled wind-wave-current conditions presents significant challenges due to strongly nonlinear hydrodynamic responses. Conventional mathematical models exhibit substantial errors under such environments, while full-scale trials are costly and offer insufficient coverage of operational scenarios, limiting reliable performance assessment. Existing grey-box approaches typically oversimplify environmental interactions by neglecting wave memory effects and lack of validation in real-sea conditions. To overcome these limitations, a novel grey-box model integrating Long Short-Term Memory (LSTM) and Deep Neural Network (DNN) architectures within the MMG framework is proposed. The LSTM branch captures wave frequency-domain characteristics and memory effects from historical data sequences, while the DNN compensates for coupled errors in hydrostatic and dynamic loads. Ablation studies and sensitivity analysis are conducted to quantify the contribution of the LSTM module. Validation through real-sea trials of a 10m planing craft in Sea States 2-3 demonstrates the model's superior accuracy in predicting trajectory, heading angle, and tactical diameter compared to theoretical models and purely data-driven model, offering a reliable foundation for the digital test and evaluation (T&E) of vessel navigation performance.
2026,48(8): 14-22 收稿日期：2025-9-3
DOI：10.3404/j.issn.1672-7649.2026.08.003
分类号：U661.7
基金项目：国家自然科学基金青年科学基金项目（52001198）
作者简介：李少楠(1997-),男,硕士,工程师,研究方向为船艇试验鉴定
参考文献：
[1] 范尚雍. 船舶操纵性[M]. 北京: 国防工业出版社, 1988.
[2] 吴晖琴. 大风浪环境舰船多工况横摇运动数学建模[J]. 舰船科学技术, 2021, 43(22): 1-3 WU H Q. Mathematical modeling of large wind and wave environment[J]. Ship Science and Technology, 2021, 43(22): 1-3
[3] ABKOWITZ M A. Measurement of hydrodynamic characteristics from ship maneuvering trials by system dentification[R]. 1980.
[4] 李冲, 章文俊, 薛宗耀, 等. 具有外源输入的船舶横摇运动NARX神经网络预测[J]. 舰船科学技术, 2022, 44(11): 63-67 LI C, ZHANG W J, XUE Z Y, et al. Improved prediction method for ship rolling motion via NARX neural network[J]. Ship Science and Technology, 2022, 44(11): 63-67
[5] KIM D, SONG S, TEZDOGAN T. Free running CFD simulations to investigate ship manoeuvrability in waves[J]. Ocean Engineering, 2021, 236(7): 109567
[6] 姜帆, 李云波, 闫凤超, 等. 基于OpenFOAM的非对称双体船操纵性预报及运动特性研究[J]. 中国造船, 2021, 62(2): 14-25 JIANG F, LI Y B, YAN F C, et al. Prediction and characteristics of maneuvering performance for asymmetric catamaran Based on OpenFOAM[J]. Shipbuilding of China, 2021, 62(2): 14-25
[7] 闫佳兵, 蔡尚峰, 赵凯. 基于CFD的船舶操纵性数值预报方法及验证[J]. 舰船科学技术, 2025, 47(10): 58-63 YAN J B, CAI S F, ZHAO K. Numerical prediction method and verification of ship maneuverability based on CFD[J]. Ship Science and Technology, 2025, 47(10): 58-63
[8] 梅斌, 孙立成, 史国友, 等. 基于单参数自调节RM-GO-LSVR的船舶操纵灰箱辨识建模[J]. 交通运输工程学报, 2020, 20(2): 88-99 MEI B, SUN L C, SHI G Y, et al. Grey box identification modeling for ship maneuverability based on single parameter self-adjustable RM-GO-LSVR[J]. Journal of Traffic and Transportation Engineering, 2020, 20(2): 88-99
[9] 姜岩, 王雪刚, 侯先瑞, 等. 深度循环神经网络在船舶操纵运动辨识中的对比研究[J]. 水动力学研究与进展A辑, 2023, 38(2): 187-194 JIANG Y, WANG X G, HOU X R, et al. Comparative study on identification of ship maneuvering motion based on deep recurrent neural network[J]. Chinese Journal of Hydrodynamics (Series A), 2023, 38(2): 187-194
[10] HAN Y, HAO L Z, SHI C, et al. Prediction of ship maneuvering motion with grey-box modelling incorporating mechanism and data[J/OL]. Ships and Offshore Structures, 2023: 1–14.
[11] RAI R, SAHU C K. Driven by data or derived through physics? a review of hybrid physics guided machine learning techniques with cyber-physical system (CPS) focus[J]. IEEE Access, 2020, 8: 71050-71073
[12] 杨波, 张炜灵, 蔡烽, 等. 风浪环境下的舰船KT操纵响应模型研究[J]. 舰船科学技术, 2020, 42(17): 99-102 YANG B, ZHANG W L, CAI F, et al. Research on KT maneuvering response model of ships in wind and wave environment[J]. Ship Science and Technology, 2020, 42(17): 99-102
[13] 韩阳, 郝立柱, 师超, 等. 规则波中船舶操纵运动预报的灰箱建模研究[J]. 中国舰船研究, 2025, 20(1): 47-57 HAN Y, HAO L Z, SHI C, et al. Prediction of ship maneuvering motion in regular wave with gray-box modelling[J]. Chinese Journal of Ship Research, 2025, 20(1): 47-57
[14] 朱冬健, 马宁, 顾解忡, 等. 波浪中船舶操纵性数值预报及自航模验证[J]. 中国舰船研究, 2015, 10(1): 76-82,96 ZHU D J, MA N, GU X C, et al. Numerical prediction and free-running model validation of ship maneuvering in waves[J]. Chinese Journal of Ship Research, 2015, 10(1): 76-82,96
[15] 王硕, 柯世堂, 张伟, 等. 风、浪、流复合作用下海上风力机基础波浪荷载分布特性研究[J]. 船舶力学, 2022, 26(5): 653-666 WANG S, KE S T, ZHANG W, et al. Research on the characteristics of wave load distribution of offshore wind turbine foundations under the combined action of wind, wave and current[J]. Journal of Ship Mechanics, 2022, 26(5): 653-666
[16] WAKITA K, MAKI A, UMEDA N, et al. On neural network identification for low-speed ship maneuvering model[J]. Journal of Marine Science and Technology, 2022, 27(1): 772-785
[17] 陈立家, 周欣蔚, 杨沛艺, 等. 面向环境不确定性的船舶操纵运动建模与预报方法[J]. 交通运输工程学报, 2024, 24(3): 279-295 CHEN L J, ZHOU X W, YANG P Y, et al. Modeling and prediction method of ship maneuvering motion facing environmental uncertainty[J]. Journal of Traffic and Transportation Engineering, 2024, 24(3): 279-295
[18] LOU J K, WANG H D, WANG J Y, et al. Deep learning method for 3-DOF motion prediction of unmanned surface vehicles based on real sea maneuverability test[J]. Ocean Engineering, 2022, 250: 111015
