【作者】 吴云峰；

【导师】 魏纳新；

【作者基本信息】 江南大学 ， 控制理论与控制工程， 2012， 硕士

【摘要】 船舶在其航行过程中将受到风、浪等环境因素干扰,产生了相互耦合的复杂的六自由度运动,尤其是在恶劣海况条件下,其运动是一个非线性随机过程,并且在船舶航行过程中不同的海区或同一海区不同时段,其统计特性有所不同,造成船舶运动的统计特性具有一定的非平稳特性。运动极短期预报对海上作业的各种海洋平台和船舶具有很重要的意义。根据运动历史数据,预报未来极短时间的运动,进而采取相应的措施,尤其是对作业时需要进行运动补偿或具有一定运动限制要求的设备。近年来,国内外关于船舶运动极短期预报的研究越来越多,时间序列法在其中受到了特别多的关注,这类算法只需要寻求出历史数据中的规律,就可以得出预报值,算法计算量小且易于实现。本文的主要工作如下:1.首先,阐述了本文的研究背景,回顾了船舶运动极短期预报的发展历程及研究现状;比较了已有的多种预报算法,并总结出各自的优缺点。2.首先,阐述了基于传统时间序列的AR模型船舶运动极短期预报算法,并将实船实测船舶运动时间序列用于预报,验证了算法的可行性,但是传统的时间序列预报算法的预报效果较差。3.在传统时间序列法的基础上引入小波多分辨率分析法,将实船所测船舶运动时间序列分解为若干层近似意义上的平稳时间序列,再使用AR模型对每层的单支重构信号进行预报,最后综合每层的预报值得到原时间序列的预报值。通过小波多分辨率分析的频带划分规则确定了最优小波分解层数的选取方法;从分析各种小波基时频特性出发,通过理论分析及仿真计算的比较,确定了最优小波基函数的选取。4.分析了经验模态分解法对非平稳信号的处理效果,针对船舶运动时间序列的非平稳性确定了其算法中相关参数的选取,运用经验模态分解法对传统时间预报算法和小波多分辨率时间序列预报算法进行算法改进。5.通过对船舶运动时间的混沌识别,确定了嵌入维数以及时间延迟的选取,使用递推最小二乘(RLS)算法对混沌Volterra级数预报模型进行参数辨识。把这类混沌时间序列的预报算法的仿真计算结果与之前所述的算法进行对比。更多还原

【Abstract】 The sea wave and ocean interfere with the ship current during its navigation process, the ship produced a complex coupled six degrees of freedom of movement, especially in the bad sea condition, its motion is a non-linear random process. And in the voyage of the same sea area at different times or in different sea areas, its Statistical properties are different. Therefore, the statistical properties of ship motion has a certain non-stationary.Short term motion prediction have a great significance for the operations of offshore platform and ship at sea. According to historical movement data, predict the future movement of a extreme short time, then take the appropriate measures, especially for case of motion compensation or have a critical motion restrictions during operating.In recent years, the study of extreme short term prediction of ship motions is becoming more and more. Among of those, Time Series Analysis Method has got the most attemion. Such algorithm only need to seek the law of the historical data, then we can get the predicted value. This algorithm computation load small, and easy to implement.The main jobs of this thesis are stated as follows:1. First, describeing the background of this study, recalling the development process and research status of the extreme short term prediction of ship motion. Compared a variety of prediction algorithms, and summarize their advantages and disadvantages.2. First, describing the traditional time series algorithms of extreme short term prediction of ship motion based on AR model, verify the feasibility of the algorithm. However, the traditional time series prediction algorithm is less effective.3. Using of multi-scale wavelet theory, non-stationary time series is decomposed into several layers of stationary time series approximately; then use time series model to predict each of layer; finally integrating each predicted layer to reconstruct the prediction of original time series. According to the original time series power spectrum analysis and the rule of the frequency band division for the wavelet multi-scale analysis, can select the best decomposed layer. According to different characteristics of wavelet basis functions, select the appropriate wavelet basis function.4. Analysis the effects of the empirical mode decomposition method for non-stationary signal processing, according to the non-stationary time series of the ship motion, select the relevant parameters in the algorithm. Use the empirical mode decomposition algorithm to improvement the traditional time series prediction algorithm and multi-resolution wavelet time series prediction algorithm.5. By the chaos identify of the ship motion, select the embedding dimension and time delay. Using the recursive least squares algorithm to identify the Volterra series prediction model parameters. Compare the simulation results of this chaotic time series prediction algorithm with the algorithm described before.更多还原