首页 | 期刊介绍 | 编委会 | 道德声明 | 投稿指南 | 常用下载 | 过刊浏览 | In English
引用本文:
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  下载PDF阅读器  关闭
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 18次   下载 0  
分享到: 微信 更多
基于二元LSTM神经网络的船动积分位移预测算法研究
张博一
上海海事大学
摘要:
摘要:在海况环境下,进行船舶运动预测时。由于惯性传感器采集系统本身的电学特性,会产生偏移误差,严重影响一般预测方法的准确性。针对这一问题,在常规LSTM神经网络的基础上,设计改良了一种二元的LSTM网络架构。在船舶运动仿真平台上进行模拟船舶升沉运动实验,并通过惯性传感系统测量仿真平台实时积分位移进行计算验证。验证统计该网络预测结果峰差值均方差0.64%,均值均方差0.42%,峰值均方差0.57%,证实该网络较常规LSTM在船舶运动预测领域具有更好的针对性和适应性,更准确的还原预测实际的船舶运动轨迹。
关键词:  船舶运动预测 LSTM神经网络 频域积分位移
DOI:10.11759/hykx20200506002
分类号:
基金项目:上海市青年科技英才扬帆计划资助
Research on ship dynamic integral displacement prediction algorithm based on binary-LSTM neural network
Zhang Boyi
Shanghai Maritime University
Abstract:
Abstract: When forecasting ship movements. Electrical errors of the inertial sensor acquisition system seriously affect the accuracy of general prediction methods. In response to this problem, based on the conventional LSTM neural network, a binary LSTM network architecture is designed and improved. The simulation of ship heave motion is carried out on the ship motion simulation platform, and the real-time integral displacement of the simulation platform is measured by the inertial sensing system for calculation and verification. Verification statistics The prediction results of the network have a peak mean square error of 0.64%, a mean square error of 0.42%, and a peak mean square error of 0.57%, which proves that the network has better pertinence and adaptability than conventional LSTMs in the field of ship motion prediction. Restore and predict the actual ship motion trajectory.
Key words:  Ship motion prediction  LSTM neural network  frequency domain integral displacement
版权所有 《海洋科学》编辑部 Copyright©2008 All Rights Reserved
主管单位:中国科学院 主办单位:中国科学院海洋研究所
地址:青岛市南海路七号  邮编:266071  电话:0532-82898755  E-mail:marinesciences@qdio.ac.cn
技术支持:北京勤云科技发展有限公司