【作者】 都延丽；

【导师】 吴庆宪；

【作者基本信息】 南京航空航天大学 ， 模式识别与智能系统， 2010， 博士

【摘要】 近空间飞行器(NSV: Near Space Vehicle)的发展涉及国家安全与和平利用空间,它是各军事大国日益关注的新型飞行器。NSV在飞行过程中呈现出大包络、多飞行状态、多任务模式的特点,并且飞行环境的特殊也使其具有严重非线性、激烈快时变、强动态不确定和强耦合的对象特征,所以NSV飞行控制系统的设计成为了一项创新而又富有挑战性的课题。围绕这一难题,本文在近空间飞行器的建模与分析、不确定环境下姿态与轨迹的非线性自适应控制等方面开展了深入的研究,主要研究成果如下:首先,建立了高超声速NSV在变化风场影响下的飞行运动模型。该模型以气流姿态角为主要状态变量,不仅包括推力矢量控制项,还包含了风场干扰项。此工作为NSV具体的飞行控制设计提供了目前国内较为完整的非线性模型。而且,本文对NSV的非线性飞行控制系统采用了系统化的设计思路。根据NSV的物理状态特点,应用串联控制结构来设计非线性自适应姿态与轨迹控制器,并依据实际飞控系统的研制程序来进行需求分析、系统设计和控制方法的选择,进一步贴近了实际系统的研制工作。然后,首次将泛函连接网络(FLANN: Functional Link Artificial Neural Network)引入到了不确定控制和飞行控制领域。FLANN比通常的多层NN计算负担小,适合实时逼近NSV受到的参数不确定和外界干扰。文中设计了FLANN干扰观测器来进行不确定和干扰的逼近,提出基于FLANN干扰观测器的非线性广义预测控制(NGPC: Nonlinear Generalised Predicitive Control)方法进行姿态控制,并首先使用了Lyapunov稳定性理论来推导FLANN的自适应控制律。之后,提出了一种新的非线性自适应控制方法——基于部分反馈FLANN的稳定自适应NGPC方法,并通过鲁棒增益的自适应调整来提高逼近动态复合干扰的精度。利用Lyapunov理论推导了能够保证误差信号一致最终有界的权值自适应律和鲁棒增益自适应律,并通过仿真验证了该方法对于存在发动机推力偏心和动态参数不确定的NSV姿态系统控制效果良好。进一步,设计B样条递归FLANN (BRFLN: B-spline Recurrent Functional Link Network),使得该网络能够学习动态的高阶非线性函数。在此基础上,针对NSV的姿态慢回路提出了PD校正的BRFLN自适应NGPC方法,并设计了完整的NSV非线性姿态控制器。对于存在高空风紊流干扰、大力矩干扰、动态参数不确定的NSV,该方法姿态控制仿真效果良好。最后,针对NSV的轨迹控制问题,在姿态控制器的基础上构造了用于稳定空速和高度的BRFLN自适应NGPC轨迹控制器。提出了一种新的群智能优化算法——改进协同微粒群算法来对控制器中BRFLN的自反馈系数进行在线学习。NSV的仿真结果表明在阵风以及姿态复合干扰存在的情况下,轨迹控制方案有效并且能良好地跟踪给定指令。更多还原

【Abstract】 The development of near space vehicle (NSV) is related to national security and peaceful use of space, and it is a new vehicle which attracts increasing attention of many military powers. The NSV in flight shows characteristics such as large flight envelope, multi-flight status and multi-tasking mode. Moreover, due to its special flight environment, the NSV also possesses objects features of serious nonlinearity, fast time variation, strong uncertainty and intense coupling.?Therefore, the flight control system design of the NSV has become an innovative and challenging issue. Around this problem, the dissertation carries out an intensive study in the NSV modeling and analysis, and the nonlinear adaptive control of attitude and trajectory system in uncertain environments. The main results are as follows:First, we establish a flight motion model of the NSV under the influence of changing wind field. Airflow attitude angles are regarded as the main state variables of this model. The state equations not only include thrust vector control, but also contain interference terms of the wind field. This work provides a more comprehensive nonlinear model at home for specific flight control design. Furthermore, a systematic design idea is applied to the design of NSV nonlinear flight control system. According to physical characteristics of the NSV, we employ series control structure to design nonlinear adaptive attitude and trajectory controllers. Also, we conduct demand analysis, system design and control methods selection based on develop procedures of actual flight control systems. This approach is further close to the actual system development work.Then, the functional link artificial neural network (FLANN) is introduced into the area of uncertain control and flight control for the first time. The FLANN bears a smaller computational burden than the usual multi-layer NN. It is suitable for the real-time approximation of parameter uncertainties and external disturbances. In this dissertation, FLANN disturbance observer (FLNDO) is designed to learn the uncertainties and disturbances, and FLNDO-based nonlinear generalized predictive control (NGPC) method is presented for the NSV attitude control. Additionally, the Lyapunov stability theory is first utilized to derive FLANN adaptive control law.Later, a new nonlinear adaptive control method is proposed. That is the stable adaptive NGPC method based on the partially feedback FLANN. And a robust control item with an adaptive gain is used to improve the accuracy of approximating dynamic lumped disturbances. The adaptive laws of network weights and robust gain are derived by the Lyapunov theory and they can ensure that system errors are uniformly ultimately bounded. Simulation results show that good attitude control effect is attained with the presented method, and the thrust misalignment disturbance and dynamic parameters uncertainties are well surpressed.Further, B-spline recurrent FLANN (BRFLN) is designed to learn high-order nonlinear functions. On this basis, PD-correction BRFLN adaptive NGPC method is proposed for the slow-loop attitude system, and then the whole attitude controller is obtained. Simulation results show satisfactory attitude control performance for high-altitude wind turbulence, large torque disturbance and dynamic parameter uncertainties.Finally, aiming at the trajectory control problem of the NSV, we construct the BRFLN adaptive NGPC trajectory controller to track the airspeed and altitude based on the attitude controller. A new swarm intelligence optimization algorithm, i.e. improved cooperative particle swarm optimizer (ICPSO), is presented to learn self-feedback coefficients of the BRFLN. Simulation results display that the trajectory control scheme is effective and the outputs can track the given instructions of the NSV which is subjected to gusts and attitude disturbances.更多还原