超恶劣海况下船舶运动简捷鲁棒自适应控制

发布时间：2018-08-08 17:35

【摘要】：为了解决特别恶劣海况(简称“超恶劣海况”,释义见附录1)现有船舶运动控制理论不适用或控制效果不理想的难题,本文以“超恶劣海况条件下的船舶运动路径跟踪控制”为选题,进行系统性的理论探索研究,期望对超恶劣海况下海难救助、不可避免的船舶大洋航行等提供理论指导。本文分别针对辨识建模算法、超恶劣海况测试平台、制导算法、船舶运动控制理论和控制工程实现5个关键点进行了深入研究,采用“鲁棒”+“自适应”+“简捷”的技术路线,即船舶运动控制算法在一定海况下具有稳健性,海况变化较大时要具有适应性,最终的控制律要形式简捷,易于实现,目标是提出一套考虑船舶控制工程需求(执行器配置约束、算法实时性)和现实船舶服役海洋环境条件的、较为完备的路径跟踪控制算法。本文首先针对船舶运动存在多变量耦合、实船操纵性试验新息(即“新信息”)长度有限等特点,将多新息辨识技术引入到船舶运动辨识建模研究中,提出了一种考虑多变量耦合机制的自组织多新息辨识算法。通过大连海事大学科研实习船“育鲲”轮开展实船试验提供数据支撑,包括旋回试验、Z形试验、非对称Z形试验,建立能够准确反映“育鲲”轮实际动态的4自由度数学模型。结合海洋环境干扰(风、海浪、海流)机理模型构建出能够描述超恶劣海洋环境的船舶运动仿真测试平台。制导和控制是船舶控制系统实现复杂自动航行任务的两大重要模块。制导方面,本文通过引入虚拟小船导引和动态切换机制,提出一种基于“动态虚拟小船”思想的制导算法,解决了已有LOS(Line-of-sight)制导算法不能直接应用于船舶路径跟踪控制这一问题。该算法能够为船舶自动航行提供更为合理的导航机制。控制方面,本文设计考虑了工程实践中存在模型高度不确定、执行器配置约束、控制工程要求算法实时性强以及超恶劣海况干扰问题,利用鲁棒神经网络控制和低频增益学习算法避免对神经网络的权重学习,通过构造在线学习的增益关联系数以镇定外界干扰和模型摄动,从而降低其对船舶动态的影响。该算法具有形式简捷、计算负载小、易于工程应用的优点。本文所有实验采用Matlab/Visual Basic混合编程实现,取得的模拟超恶劣海况下的实验结果验证了所提出路径跟踪控制设计方法的有效性。该研究对提高船舶海上安全保障、推进船舶国产化装备研制具有现实意义,为实现高效航运奠定了重要理论基础。
[Abstract]:In order to solve the problem that the current theory of ship motion control is not applicable or the control effect is not ideal, in order to solve the problem of especially bad sea conditions (referred to as "ultra bad sea conditions", which is explained in Appendix 1). Based on the topic of "tracking and controlling the ship's moving path under the condition of ultra bad sea conditions", this paper makes a systematic theoretical exploration and study, hoping to provide theoretical guidance for the rescue of shipwrecks in the extreme bad sea conditions and the inevitable navigation of ships in the ocean. In this paper, five key points, such as identification modeling algorithm, ultra-bad sea condition test platform, guidance algorithm, ship motion control theory and control engineering realization, are studied in depth. Adopting the technical route of "robust" and "adaptive" and "simple", that is, the ship motion control algorithm is robust under certain sea conditions, and should be adaptive when the sea condition changes greatly, and the final control law should be simple and easy to realize. The aim of this paper is to propose a complete path tracking control algorithm which takes into account the requirements of ship control engineering (actuator configuration constraints, real-time algorithm) and the actual marine environment conditions of ship service. Aiming at the characteristics of multivariable coupling of ship motion and limited length of real ship maneuverability test innovation (i.e. "new information"), this paper introduces multi-innovation identification technology into ship motion identification modeling research. A self-organizing multi-innovation identification algorithm considering multivariable coupling mechanism is proposed. Through carrying out the actual ship test on the scientific research practice ship "Yukun" of Dalian Maritime University, including cycle test and asymmetric Z-shape test, a four-degree-of-freedom mathematical model which can accurately reflect the actual dynamics of "Yukun" is established. Combined with the mechanism model of ocean environment disturbance (wind, wave, current), a ship motion simulation test platform can be constructed to describe the ultra-bad marine environment. Guidance and control are two important modules of ship control system to realize complex automatic navigation task. In the aspect of guidance, by introducing virtual boat guidance and dynamic switching mechanism, this paper proposes a guidance algorithm based on the idea of "dynamic virtual boat", which solves the problem that the existing LOS (Line-of-sight) guidance algorithm can not be directly applied to ship path tracking control. The algorithm can provide a more reasonable navigation mechanism for automatic navigation of ships. In the aspect of control, this paper considers the problems of model height uncertainty, actuator configuration constraint, real-time control algorithm and super bad sea condition interference in engineering practice. Robust neural network control and low-frequency gain learning algorithm are used to avoid the weight learning of neural network. The gain correlation coefficient of on-line learning is constructed to stabilize the external disturbance and model perturbation so as to reduce the influence on ship dynamics. This algorithm has the advantages of simple form, small computational load and easy engineering application. All experiments in this paper are implemented by Matlab/Visual Basic hybrid programming. The experimental results obtained under the ultra-bad sea conditions verify the effectiveness of the proposed path tracking control design method. The research is of practical significance for improving the safety of ships at sea and promoting the development of home-made equipment for ships, and lays an important theoretical foundation for the realization of efficient shipping.
【学位授予单位】：大连海事大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：U661.3

【相似文献】