【作者】 曹莹；

【导师】 王岩；

【作者基本信息】 哈尔滨工业大学 ， 控制科学与工程， 2010， 硕士

【摘要】 卫星姿态控制系统是一个耦合的不确定非线性系统。在轨运行的卫星不可避免地受到模型参数不确定性和各种干扰力矩的影响,这些不确定性的存在使挠性卫星大角度姿态机动的控制问题进一步复杂化。因此,为了完成姿态控制任务,需要所设计的控制律具有较高的鲁棒性。本文就是在这种背景下,从理论和应用两个方面对卫星姿态控制系统的控制算法进行了深入的研究。主要完成了以下几个方面的工作:首先,滑模控制有很多优点,如鲁棒性好、计算量小、实时性好、响应速度快等。因此,用滑模控制方法来控制挠性卫星姿态,用饱和函数代替符号函数来消除抖振现象。其次,用滑模变结构进行挠性卫星姿态机动控制,用神经网络补偿不确定性,提高系统鲁棒性。采用传统的小脑神经网络来逼近不确定性,此网络不容易产生局部极小现象。针对传统小脑神经网络实时性差,泛化能力不够好的缺点,研究了高斯基函数的小脑神经网络,运用高斯基函数代替了传统小脑神经网络量化的0或者1。针对高斯基函数小脑神经网络计算量大,应用不方便,研究了基于超立方体子空间的快速学习算法,大大提高了网络的学习速度。以上的网络都需要知道输入的具体范围,这样对网络的应用有一定的限制,自组织小脑神经网络不需要知道输入范围,可根据输入值自动更新节点数和权值。这样对于挠性卫星姿态控制系统参数的变化能通过设计来满足精度要求。更多还原

【Abstract】 Satellite attitude control system is a coupled uncertain nonlinear system. For any on-orbit satellite, it is inevitable to be influenced by some kinds of uncertain parameters and disturbance torques, which makes the attitude control problem further complicated. Therefore, to accomplish attitude control mission, it is necessary to design attitude control laws with high robustness. On this background, this thesis investigated attitude control algorithms for satellite attitude control system in detail, from both theoretical and applicable aspects, and applied the proposed control schemes to certain satellite control system. The main contents of this thesis are as follows.First, Sliding mode control has many advantages, such as high robustness, easy calculation, good real-time characteristic, quick response characteristic. The paper use this theory to realize the attitude maneuver control of flexible satellite,using saturation function replace sign function to Eliminate Buffeting Phenomenon.Second, the paper use sliding mode to control the satellite, using neural networks to compensate the uncertainty, which is proved by Lyapunov stability theory.The paper use tradition CMAC neural network method to approach uncertain function, which can solve the local minimum phenomenon. However, the tradition CMAC has bad real-time and generalization characteristic, soGaussian function CMAC, which can quantified as a continuous gaussian function, is proposed to replace the tradition CMAC. Despite of this, gaussian function CMAC is complex to calculation, a fast algorithm CMAC is a new way to improve computing speed.All of the above neural networks, the range of the input signal must be awared of in advance, which will limit the application of the network. However, it is not necessaryto know the range of the input signal using self-organization CMAC neural network, which can update of structure and weights automatically. It is more convenient and intelligent.So, for the changes in parameters, this flexible satellite attitude control system can be designed to meet the accuracy requirements.更多还原