基于马尔可夫蒙特卡洛法的系统辨识方法研究及应用

提出了基于马尔可夫蒙特卡洛(MCMC)的贝叶斯辨识方法,以解决高超声速飞行器系统辨识中复杂动力学模型转换为简单或稀疏模型所带来的不确定性问题,以及存在的训练数据大和积分难处理的问题。该方法将数据退火算法引入MCMC中,不仅解决了MCMC易陷入局部最优的问题,并且将数据退火与“高信息训练数据”的概念相结合,能够以较低的计算成本分析大数据集。此外,该方法可以对参数估计过程中存在的不确定性进行量化,获得未知参数的最优估计值。通过仿真实验,验证了提出的系统辨识方法的有效性,辨识出的模型能够有效应用于控制器设计之中,并获得较好的控制效果。     

火箭整流罩锥壳夹层结构不确定性轻量化设计

为分析运载火箭整流罩锥壳夹层结构不确定性对结构热稳定性的影响并指导结构的轻量化设计,建立整流罩前锥段夹层圆锥壳模型,并建立温度场模型,据此对圆锥壳开展热稳定性分析,推导力热联合载荷作用下整流罩前锥段夹层结构失稳临界轴压。在此基础上,针对主要不确定性因素,开展灵敏度分析并建立区间不确定性优化模型,采用区间可能度方法将其转化为确定性问题,并采用遗传算法-区间分析算法（GA-CIAM）实现结构优化设计。计算结果表明:考虑气动力/热载荷及材料参数不确定性影响,对整流罩前锥段结构开展优化设计,在满足设计要求的前提下,有效实现结构轻量化。      

2D雷达组网中目标高度估计误差的Cramér-Rao限

 在由2坐标雷达组成的雷达网中,推导了目标高度估计误差的CRLB(Cram&;#225;r-Rao限),并通过不同条件下的数值计算得到了一些结论。结果表明,目标高度估计误差的CRLB既与雷达的测角误差有关,也与目标和2个雷达站形成的夹角有关系,雷达配置在不同的高度上有利于目标高度估计的收敛性。这些结论对于2坐标雷达组网以及雷达网中的传感器管理具有指导意义。     

基于Jerk输入估计的MCS模型及非线性跟踪算法

针对强机动目标跟踪问题,基于当前统计(CS,Current Statistical)模型、改进输入估计 (MIE,Modified Input Estimation)和无迹强跟踪滤波器,提出了一种新的自适应目标跟踪算法.该算法引入Jerk输入估计改进了当前统计模型的状态方程和机动加速度方差调整方法,利用改进的无迹强跟踪滤波器实现了状态协方差、状态噪声协方差和机动频率的联合自适应.在没有加速度先验知识的情况下,能够实时准确跟踪目标连续强机动、匀加速机动和匀速运动状态.仿真实验表明:相比CS模型无迹滤波算法、CS模型无迹强跟踪算法和交互多模型算法,该算法在对目标强机动的适应性、跟踪精度和对突变状态跟踪的收敛性方面都有更好的性能.     

基于极点配置的真空热试验温度轨迹自适应补偿控制

为了在航天器真空热试验中对试件温度按指定轨迹精确控制,通过变量代换对真空冷背景环境下的对象模型进行稳态线性化近似处理;通过系统辨识技术对广义对象的时滞和模型参数进行了联合辨识;依据对象估计模型,按照极点配置方法设计了自适应控制律,使闭环系统具有期望的闭环稳定性;根据估计模型的延迟和闭环期望特性设计了自适应补偿器,从而提高了控制系统对给定输入的跟踪性能.运用该方法在模拟太阳翼的温度轨迹跟踪试验中取得了满意的控制效果.     

基于卡尔曼滤波和主元分析的发动机性能衰退

在给定测量参数基础上,通过非线性估计得到对应离散时刻的部件性能参数,利用数据插值方法扩充测量参数。利用主元分析方法对这些测量参数进行分解,提取反映性能衰退的成分,以此构造反映性能变化的统计量,假定该统计量服从,分布,对性能衰退情况做进一步分析,得到在10000时间点附近性能衰退最为严重,已超过阀值,与前面参数估计结果相符。用部件性能衰退分析方法对发动机性能状况监控,为及时采取维护措施提供了一定的指导作用。     

H∞ output tracking control for flight control systems with time-varying delay

For flight control systems with time-varying delay, an H∞ output tracking controller is proposed. The controller is designed for the discrete-time state-space model of general aircraft to reduce the effects of uncertainties of the mathematical model, external disturbances, and bounded time-varying delay. It is assumed that the feedback-control loop is closed by the communication network, and the network-based control architecture induces time-delays in the feedback information. Suppose that the time delay has both an upper bound and a lower bound. By using the Lyapu- nov-Krasovskii function and the linear matrix inequality （LMI）, the delay-dependent stability criterion is derived for the time-delay system. Based on the criterion, a state-feedback H∞ output tracking controller for systems with norm-bounded uncertainties and time-varying delay is presented. The control scheme is applied to the high incidence research model （HIRM）, which shows the effectiveness of the proposed approach.     

Research on uncertainty in measurement assisted alignment in aircraft assembly

Operations in assembling and joining large size aircraft components are changed to novel digital and flexible ways by digital measurement assisted alignment.Positions and orientations（P＆O）of aligned components are critical characters which assure geometrical positions and relationships of those components.Therefore,evaluating the P＆O of a component is considered necessary and critical for ensuring accuracy in aircraft assembly.Uncertainty of position and orientation（U-P＆O）,as a part of the evaluating result of P＆O,needs to be given for ensuring the integrity and credibility of the result;furthermore,U-P＆O is necessary for error tracing and quality evaluating of measurement assisted aircraft assembly.However,current research mainly focuses on the process integration of measurement with assembly,and usually ignores the uncertainty of measured result and its influence on quality evaluation.This paper focuses on the expression,analysis,and application of U-P＆O in measurement assisted alignment.The geometrical and algebraical connotations of U-P＆O are presented.Then,an analytical algorithm for evaluating the multi-dimensional U-P＆O is given,and the effect factors and characteristics of U-P＆O are discussed.Finally,U-P＆O is used to evaluate alignment in aircraft assembly for quality evaluating and improving.Cases are introduced with the methodology.     

超模式斜采样遥感图像超分辨复原方法

超模式斜采样成像技术通过将两排探测器阵列倾斜成像提高相机空间分辨率。由于不满足香农采样定理．图像中出现混叠。为了降低混叠对遥感图像超分辨复原的影响,文章提出了一种结合最佳倒易晶胞的超分辨复原方法。该方法根据超模式斜采样频域对偶网格建立相应的混叠模型．通过权重函数衡量噪声和混叠在频谱中的分布,得到噪声、混叠最小最佳倒易晶胞;并基于Huber—MRF（Markov Random Field）图像先验模型建立了耦合最佳倒易晶胞的MAP（Maximuma Posterior）超分辨复原模型。实验表明,该方法降低了超模式斜采样成像中混叠对复原的影响,获得了良好的超分辨复原效果。