【作者】 李雯；

【导师】 戴金海；

【作者基本信息】 国防科学技术大学 ， 航空宇航科学与技术， 2007， 博士

【摘要】 网络控制系统(NCS)作为一种新型的控制系统,以其优秀的系统结构较好地满足了现代飞行器智能性和自主性的要求,代表了飞行器控制系统的发展趋势。论文针对NCS的共性问题和飞行器应用的特有问题,对飞行器网络控制系统(FV-NCS)进行了深入的研究。首先,选择了适用于飞行器系统的控制网络。通过对1553B总线、CAN总线和工业以太网的分析和对比,最终选择CAN总线来组建FV-NCS。通过对CAN总线在飞行器系统的应用实例进行分析,证明了飞行器应用CAN总线是可行的。通过仿真实验和理论分析,研究了CAN网络控制系统(CAN-NCS)的时延特性。讨论了NCS时延的两种组成分析方法,并揭示了两者的关系;利用数字仿真工具TrueTime对NCS进行仿真,探讨了网络协议、网络负载等因素对消息时延的影响;针对CAN-NCS,运用实时系统任务调度方法,给出了CAN总线最差情况消息等待时间的连续和离散调度公式,确定了CAN总线数据帧的最长传输时间计算公式,总结了CAN-NCS时延的一般特性:利用TTCAN改进了CAN-NCS时延的实时性,并分析了TTCAN飞行器应用的可行性。通过输出响应和闭环带宽,研究了CAN-NCS时延对飞行器控制系统的影响。研究表明,CAN-NCS时延基本不影响飞行器的质心运动和姿态运动,但对带宽较高的伺服机构将产生明显影响。论文以直流伺服电机作为这类时延敏感对象的代表,开展后续控制器的研究,确保复杂、多回路的FV-NCS的整体性能。对短时延NCS,针对四种不同情况,建立了形式统一的广义被控对象模型和包括控制器在内的闭环系统模型,为采用通用方法分析、设计NCS奠定了基础。研究了FV-NCS的自适应控制。设计了误差积分型LQG控制器作为基本控制器,解决了标准LQG控制器无法跟踪给定输入的问题;利用TrueTime对基本控制器控制的直流伺服电机NCS进行了仿真实现,研究了采样周期和网络时延对系统性能的影响;对固定的网络时延,利用状态增广法,设计了参数固定的LQG控制器;对时变的网络时延,设计了参数自适应LQG控制器,提出了合理增大控制器参数的更新周期、自动切换控制策略等方法,较好地解决了系统响应严重振荡问题;对传感器数据丢包和时变的网络时延,通过引入事件/时间混合驱动方式和丢包处理算法,设计了带数据丢包处理的参数自适应LQG控制器。利用TrueTime对上述三种控制器进行了仿真实验,实验结果表明,对各自针对的情形,三种控制器均获得了很好的控制效果。研究了FV-NCS的鲁棒控制。针对一般的系统矩阵,建立了统一的CAN-NCS鲁棒控制模型,提出了一种不确定矩阵界值的实用计算方法,满足了鲁棒控制模型对不确定矩阵的要求;对直接状态反馈,推导了NCS鲁棒稳定和H_∞控制的充分条件,通过适当变换,获得了线性矩阵不等式(LMI)形式的约束;对带观测器状态反馈,针对参数矩阵中包含矩阵变量的问题,提出了矩阵变量逆阵的近似线性化方法,确定了NCS鲁棒稳定和H_∞控制的LMI求解算法;针对直流伺服电机NCS,利用MATLAB的LMI工具箱获得了各可行解,并在TrueTime下进行了仿真实验,实验结果表明,鲁棒稳定的可行解确保了闭环系统稳定,H_∞控制的可行解在确保闭环系统稳定的前提下,对外部干扰具有较好的抑制作用。研究了FV-NCS的增益调度控制。探讨了对原控制器进行输出增益调度的一般方法;以误差积分型LQR控制器为原控制器,研究了固定网络时延的增益调度方法,提出了多级寻优策略,实现了最优增益的快速确定,改进了代价函数,解决了以误差平方和为代价函数时,最优增益的系统性能并非最优的问题;通过在线估计网络时延和动态调度增益系数,实现了时变网络时延的增益调度控制;利用TrueTime进行了仿真实验,实验结果表明,与原控制器相比,增益调度控制的系统性能有明显改善,可以使原控制器以尽可能小的代价较好地适应网络环境。更多还原

【Abstract】 As a new kind of control system, networked control systems (NCS) satisfies intelligent and autonomous requests of modern flight vehicle by the good system architecture, and represents the future development of flight vehicle control system. This thesis focuses on NCS of flight vehicle (FV-NCS) both the common problems of NCS and the special problems for flight vehicle application.The control network which is adapted to flight vehicle control system is studied. 1553B bus、CAN bus and Ethernet are analyzed and compared, and CAN bus is selected to build FV-NCS. The examples of CAN bus being applied to flight vehicle control system indicate the feasibility of flight vehicle system applying CAN bus.The delay characteristic of NCS based on CAN (CAN-NCS) are researched by simulation experiments and theoretical analysis. Two methods of analyzing the components of NCS delay are discussed and their relationship is described. NCS is simulated by TrueTime and message delay characteristic which is affected by network protocol, network load and other factors is discussed. For CAN-NCS, the formula of continuous scheduling and discrete scheduling for the worst case message waiting time of CAN are given by methods of real-time system task scheduling, the formula of the longest transmission time of CAN data frame is defined, the generic characteristic of CAN-NCS delay is obtained. Real-time capability of CAN-NCS is improved by TTCAN, the feasibility of flight vehicle system applying TTCAN is analyzed.The influences of CAN-NCS delay to flight vehicle control system are researched by output response and bandwidth of closed loop system. It is discovered that CAN-NCS delay can not affect centroid movement and attitude movement of flight vehicle, but CAN-NCS delay can affect many servo mechanisms which always have higher bandwidth. As the representation of these type of plants which are sensitive to CAN-NCS delay, DC motor is used as the control plant in this thesis to develop the controller study, in order to ensure the whole performance of complex and multiple loops of FV-NCS. As the basic of analyzing and designing of NCS by generic methods, the uniform augmented plant model and closed loop system model of short delay NCS are built for four different cases.Adaptive control of FV-NCS is studied. Error-integral LQG controller is designed as the basic controller, in order to solve the problem that standard LQG controller can not track the giving input. The simulation of DC motor NCS with basic controller based on TrueTime is realized, the influences by sampling period and network delay to control system performance are analyzed. For fixed network delay, fixed parameter LQG controller is designed by state argumentation. For time-varying network delay, adaptive parameter LQG controller is designed. The methods of increasing the update period of controller parameter and automatically switching control strategies are put forward and the problem of serious oscillation of system response is solved . Considering sensor package loss and time- varying network delay, adaptive parameter LQG controller with package loss managing is designed by mixed event/time driving mode and arithmetic of package loss managing. The simulation experiments for the three kinds of controller are made based on TrueTime, the results indicate that the three kinds of controller all achieve good control effect for the certain case.Robust control of FV-NCS is investigated. Uniform robust control model of CAN-NCS is built for a generic system matrix. A calculate method of the bound value for uncertain matrix is brought forward to satisfy the demand for uncertain matrix in the robust control model. For direct state feedback, sufficient conditions of stability and H_∞control of NCS are deduced, the restricts of LMI form are obtained by appropriate transformation. For state feedback with observer, because parameter matrix contains matrix variables, the approximate linearization method of inverse matrix of matrix variable is presented, and the arithmetic of LMI form for stability and H_∞control of NCS are given. For DC motor NCS, the feasible solutions are obtained by LMI toolbox of MATLAB, the simulation experiments are made based on TrueTime. The results indicate that the feasible solutions of robust stability can guarantee the closed loop system stable and the feasible solutions of H_∞control can restrain the external disturbances effectively.Gain scheduling control of FV-NCS is studied. The generic method for original controller to schedule output is discussed. Error-integral LQR controller is used as the original controller, gain scheduling method for fixed network delay is researched, the multilevel searching strategy is put forward and the optimal gain is quickly found, cost function is modified and the problem is solved that system performance of the optimal gain is not the best when sum of error square is the cost function. Gain scheduling control for time-varying network delay is realized by on-line delay estimation and dynamic scheduling gain coefficient. The simulation experiments are made based on TrueTime, the results indicate that the system performances are improved by gain scheduling control comparing to original controller and gain scheduling control can make original controller to adapt the network environment with the possible little expense.更多还原