华中科技大学学报（自然科学版）2025，Vol.53Issue(4)：132-137,6.DOI:10.13245/j.hust.250504
基于深度学习的舰船姿态预估的可视化研究
Visualization research on prediction of ship attitude based on deep learning
摘要
Abstract
Aiming at the problem that the drastic change of motion attitude would greatly reduce the safety of carrier-based aircraft when the ship was sailing in the actual sea conditions,a composite prediction model was formed by introducing long short-term memory neural network(LSTM)and K-means clustering algorithm(K-Means)into the modular neural network(MNN).First,the model was trained and the best parameters were saved based on the simulation values generated by Fortran software.Then,the parameters were invoked and the attitude was predicted based on the experimental data of ship model.The minimum predicted loss value could reach 1×10-5 order of magnitude,and the maximum fitting coefficient could reach 0.98.Meanwhile the MNN-KMEANS-LSTM model was compared with radial basis function and gated cyclic unit neural network to analyze the prediction performance of different models under strong nonlinear non-stationary sea state.Finally,with the help of third-party library of Python(PyQt),the ship motion attitude display platform was developed to carry on the theoretical research results.Based on the three functions of the software interface,the model adaptivity meets the online calculation requirements and realizes human-computer interaction,which will provide a pre-judgment reference for ship staff to cooperate with carrier-based aircraft takeoff and landing.关键词
舰船运动/长短期记忆神经网络/模块化神经网络/PyQt第三方库/在线计算Key words
ship motion/long short-term memory neural network/modular neural network/PyQt third-party library/on-line computation分类
交通工程引用本文复制引用
陈占阳,王振宇,崔海鑫,黎胜..基于深度学习的舰船姿态预估的可视化研究[J].华中科技大学学报（自然科学版）,2025,53(4):132-137,6.基金项目
山东省自然科学基金面上项目(ZR2024ME139) （ZR2024ME139）
航空科学基金资助项目(2024M074189001) （2024M074189001）
工业装备结构分析国家重点实验室开放基金资助项目(GZ23112). （GZ23112）
