基于固定网格小波神经网络的不规则波中船舶横摇运动在线预报  

Online Prediction of Ship Roll Motion in Irregular Waves Using a Fixed Grid Wavelet Network

作　　者：黄柏刚 邹早建[1,2] HUANG Bai-gang;ZOU Zao-jian(School of Naval Architecture,Ocean and Civil Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;State Key Laboratory of Ocean Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学船舶海洋与建筑工程学院,上海200240 [2]上海交通大学海洋工程国家重点实验室,上海200240

出　　处：《船舶力学》2020年第6期693-705,共13页Journal of Ship Mechanics

基　　金：国家自然科学基金资助项目(51779140)。

摘　　要：船舶横摇运动预报对于船舶安全与作业非常重要。本文应用固定网格小波神经网络在线预报不规则波中的船舶横摇运动。该固定网格小波神经网络由离散的小波激活函数组成,其结构和参数可以基于滑动数据窗在线调整;在每一个滑动数据窗,误差下降比判据被用来从小波函数库中选择重要的小波函数项来构建小波神经网络模型,直到该模型可以较好地表达所研究的非线性系统,获得的模型一般比较简洁。预报结果表明,仅仅几个小波函数项就可以很好地捕捉到不规则波中船舶横摇运动的非线性动力学内在特性,这不仅展示了小波函数很强的非线性表达能力,也证实了所采用的建模方法对于预报船舶在不规则波中的横摇运动的有效性。Ship roll motion prediction is important for ship safety and operation.A fixed grid wavelet network is applied for online prediction of ship roll motion in irregular waves.The proposed network is composed of discrete wavelet activation functions.The structure and parameters of the wavelet net⁃work can be adjusted online based on a sliding data window.In every sliding data window,the error re⁃duction ratio criterion is used to select the significant wavelet functional components in the library of wavelet terms until the selected wavelet components can give a good representation of the nonlinear system.The obtained model is concise.The prediction results show that the intrinsic characteristics of the nonlinear dynamics of ship roll motion in irregular waves can be captured by only several wavelet components,indicating the strong nonlinear representation ability of wavelet and demonstrating that the modeling method can effectively predict ship roll motion in irregular waves.

关 键 词：船舶横摇运动 不规则波 在线预报 固定网格小波神经网络 滑动数据窗 

分 类 号：U661.32[交通运输工程—船舶及航道工程]