【作者】 李伟；

【导师】 刘正江；

【作者基本信息】 大连海事大学 ， 交通信息工程及控制， 2014， 博士

【摘要】 随着造船业及航运领域的蓬勃发展,船舶也逐步趋于大型化、高速化及自动化,对船舶操控性的要求也越来越高,船舶自动舵的研发受到了空前的重视。船舶在海上航行时,由于航速、装载状态等的变化以及外界环境干扰的影响,致使船舶控制系统具有强不确定性、非线性及不稳定性。因此,研制控制性能优异、易于工程实现的非线性船舶航迹自动舵显得尤为重要。本文针对一类单输入单输出不确定非线性系统,基于Lyapunov稳定性理论,人工神经网络(Artificial Neural Networks, ANNs)逼近器、反步技术(Backstepping)、动态面控制技术(Dynamic Surface Control, DSC)和最少学习参数(Minimal Learning Paramaters, MLP)算法,提出了一种新的的自适应神经网络控制方法。将神经网络算法用于逼近控制系统中未知的不确定函数,所得到的闭环系统是半全局一致最终有界的。所提出的算法最显著的特点是：(1)在控制器的设计过程中,当采用反步(Backstepping)方法时,会引发“计算量膨胀”问题,本文引入了DSC技术,其较好地处理了设计过程中的“计算量膨胀”问题。(2)采用人工神经网络技术去逼近控制系统中的未知不确定函数时,针对自适应神经网络算法中易产生“维数灾难”的问题,本文引入了MLP算法解决了这个问题。将MLP与DSC结合,旨在同时解决“维数灾难”及“计算量膨胀”问题,克服控制器设计时可能存在的控制器奇异值问题,从而保证闭环系统的稳定性。另一方面,由于船舶运动呈现非线性、强不确定性及复杂性,将提出的单输入单输出非线性系统自适应神经网络控制算法应用到船舶直线航迹保持控制系统中,本文共设计了三种具有鲁棒性、自适应性、可靠性和实时性的船舶航迹保持控制算法；三种控制算法分别针对船舶航迹保持控制系统中的自适应学习参数进行了调整,Matlab仿真实验验证了控制器的控制性能。最后改变控制系统模型参数,并与传统的PD舵控制效果进行对比研究,验证本文提出的控制算法的鲁棒性、自适应性,同时具有计算量小,易于工程实现的特点。最后将船舶直线航迹自动控制技术应用于船舶通航安全模拟,将航迹自动控制模块嵌入船舶操纵模拟器,提出了一种基于自动船舶操纵模拟器的船舶通航安全模拟方法,将其应用于船舶在受限水域的通航安全评价研究,弥补了人工船舶操纵模拟器受人为因素影响的缺陷,极大改进了船舶操纵模拟器技术,为船舶安全操纵提供科学、客观的参考决策,进而开发出船舶在各种港口、航道及河道等限制水域内的通航安全智能评估方法,为限制水域内船舶的安全操纵提供客观、可行的参考策略；为船舶安全、快速通过限制水域提供指导,以切实提高船舶通航的安全性和效率；为航道整治及拟建和新建港航工程设施的通航安全论证和评估提供科学、有效的通航模拟和评估方法。更多还原

【Abstract】 With the vigorous development of the shipbuilding and shipping area industry, the developments of ship also gradually become large-scale, high speed and automation, the requirement of ship handling is becoming more and more higher, research and development of ship autopilot has received the unprecedented attention. When the ship is navigating in the ocean, due to the change of speed and loading condition and also the influence of external environment disturbance, ship control system shows strong uncertainty, nonlinearity and instability. Therefore, the nonlinear ship-tracking autopilot which has the control performance of excellent and easy to realize project implementation is particularly important.Based on the control theories and methods, such as Lyapunov stability theory, Artificial Neural Networks, Backstepping technique, Dynamic Surface Control technology and Minimal Learning Paramaters algorithm, this paper proposes a new class of self-adaptive neural network control algorithm for the non-linear single input and output system. Artificial Neural networks is used to approach the unknown uncertain functions in the control system, and the closed-loop system can achieve semi-globally uniformly ultimately bounded. The most significant feature of the algorithm are:(1) This paper introduces DSC technology which can slove the problem of "explosion of complexity" caused by the traditional Backstepping technique during the process of the controller design.(2) When employing neutral networks to approximate the uncertain functions in the control system, sometimes it will cause the problem of "dimension curse" which will enlarge the designing workload. For this reason, MLP algorithm is introduced to cope with this problem. For the sake of solving the above-mentioned problems simultaneously, MLP algorithm and DSC method is combined. In this way it can not only use less parameters and reduce computational load, but also avoid the possible existing problem of controller singularity. In addition, the stability of the closed-loop system is guaranteed and the tracking error converges around zero.On the other hand, the motion of ship shows nonlinearity, strong uncertainty and complexity, for these reason the proposed self-adaptive neural network control algorithm of the non-linear single input and output system in this paper is being applied to the ship liner-track keeping control system.Three kinds of linear-track keeping control algorithm with good robustness, self-adaptively, reliability and real-time are put forward finally.. The proposed three control algorithms are focused on adjusting the adaptive learning parameters in the ship, track keeping control system, so the number of the parameters in each subsystem are different. Matlab simulation results validate the effectiveness of the proposed algorithm. To further illustrate the performance of the proposed algorithm with tradition PD autopilot, change the parameters of ship’s control system, through comparing the simulation results, we find that the proposed algorithm has the property of adaptive, robustness, simple, and easy to be implemented in applications.Finally, a new assessment technique for ship navigation safety in the constrained waters is proposed based on ship’s linear-track keeping control technique. By embedding the linear-track keeping control module in the ship handling simulator (SHS), a traffic simulation method based on autoSHS for the safety of ship navigation is finally proposed. By applying this simulation method to the evaluation of the safety of the ship navigating in restricted water, the defect of handSHS which is affected by human factors can be make up, and the SHS technology can be improved greatly. A scientific and objective reference decision can also be gained due to the safety of the ship manipulating. Intelligent navigation safety assessment methods for ships in restricted waters, such as a variety of ports, channels and rivers will be developed further. Meanwhile, objective and feasible reference strategies will also be given for the safety of ship manipulation in restricted waters. What’s more, safety and efficiency of the ship traffic will be improved practically by providing guidance how to behave safely and speedy; Scientific and effective navigation simulation and evaluation methods will be provided for navigation safety demonstration and evaluation of channel improvement projection and the new port facilities.更多还原