船舶航行速度直接影响船舶航行安全与路径规划效果，为此建立基于改进神经网络的船舶航行速度估计数学模型，提升船舶航行速度估计效果。利用位置编码机制，编码处理风速与波高等船舶航向速度相关数据；在编解码卷积神经网络内，引入自注意力机制，得到改进编解码卷积神经网络；利用编码后的船舶航速训练数据，训练改进神经网络，建立船舶航行速度估计数学模型；在该模型内，输入编码后的船舶航速测试数据，通过自注意力机制，提取船舶航速数据特征，并结合自注意力蒸馏剔除冗余特征；通过全连接层处理船舶航速数据特征，输出船舶航行速度估计结果。实验证明，该模型可有效提取船舶航速数据特征；该模型可精准估计船舶航行速度；在不同浪向下，该模型船舶航行速度估计的决定系数均较高，即估计精度较高。
Ship sailing speed directly affects ship sailing safety and route planning effect. Therefore, a mathematical model of ship sailing speed estimation based on improved neural network is established to improve ship sailing speed estimation effect. The position coding mechanism is used to encode and process the relevant data of wind speed and ship heading speed with wave height. In the CODEC convolutional neural network, self-attention mechanism is introduced to improve the CODEC convolutional neural network. Using the coded ship speed training data, the improved neural network is trained and the mathematical model of ship speed estimation is established. In this model, the ship speed test data after coding is input, the ship speed data features are extracted by self-attention mechanism, and the redundant features are eliminated by self-attention distillation. Through the full connection layer, the ship speed data characteristics are processed, and the ship speed estimation results are output. Experimental results show that the model can effectively extract the characteristics of ship speed data. The model can accurately estimate ship speed. Under different waves, the determination coefficients of the model are higher, that is, the estimation accuracy is higher.
2024,46(4): 70-73 收稿日期：2023-10-30
DOI：10.3404/j.issn.1672-7649.2024.04.014
分类号：U661.31
作者简介：姚光文(1967-),男,硕士,讲师,研究方向为数学建模
参考文献：
[1] 王均刚, 丁惠倩, 胡柏青. 基于滑动窗口PSO-LSSVR的船舶轨迹预测模型[J]. 武汉理工大学学报, 2022, 44(12): 35-43,59.
[2] 贾雨薇, 邵忻. 基于多蚁群系统的多无人机反侦察航迹规划[J]. 舰船电子对抗, 2023, 46(3): 14-19.
[3] 张青山, 陈伟民, 杜云龙, 等. 基于实尺度船舶自航数值模拟的航速预报[J]. 中国造船, 2022, 63(5): 11-21.
[4] 朱晓晨, 尹奇志, 赵福芹, 等. 基于LightGBM的船舶航速预测模型[J]. 大连海事大学学报, 2023, 49(1): 56-65.
[5] 张阳, 高曙, 何伟, 等. 基于多模型融合的内河船舶航行轨迹预测方法[J]. 中国机械工程, 2022, 33(10): 1142-1152.
[6] 刘长德, 顾宇翔, 张进丰. 基于小波滤波和LSTM神经网络的船舶运动极短期预报研究[J]. 船舶力学, 2021, 25(3): 299-310.
