【作者】 姚敏；

【导师】 赵敏；

【作者基本信息】 南京航空航天大学 ， 测试计量技术及仪器， 2008， 博士

【摘要】 无论是微小卫星的单星运行还是编队飞行或者组成星座系统,都对星务系统的智能性和自主性提出了越来越高的要求。本文以南航正在研制的“天巡一号”微小卫星为研究对象,对其星务系统设计中任务调度分配、温度测量冗余设计、遥感图像压缩编码、故障诊断等关键技术进行了研究。首先,运用“合一-阈值”启发式调度方法对星务系统任务进行了分配,针对“合一-阈值”启发式调度算法处理任务数量有限、难以获得任务分配最优解的缺点,研究了基于遗传算法的任务调度分配方法,得到了任务分配的最优解。为了实现星务系统任务的动态调度,并且在调度中充分考虑任务的偏序关系、不同任务执行条件,本文提出了一种基于模糊神经网络的智能化动态任务调度方法。该方法根据任务的偏序关系对任务进行分类,根据当前的光照等条件自动决策当前的任务需求,实现了任务的动态调度。根据微小卫星温度遥测点多,仅用硬件进行冗余备份设计会给微小卫星带来重量、功耗和体积的增加等问题,本文提出一种基于数据融合的冗余设计方法。根据传感器的分布位置,将传感器划分为多个组,同组传感器通过数据融合的方法相互提供参考,给出每个传感器测量值的可信度。在可信度较低时,对其测量值不予使用,这种冗余设计方法提高了温度测量的可靠性。卫星在正常运行中,为了减少下传通信数据量,简化遥测监控任务,需要通过多个传感器给出温度区域的估计值。本文应用基于算术平均和分批估计的方法和基于Bayes参数估计的方法对区域温度进行估计,得到了比算术平均更高的估计精度。特别是在存在故障传感器的情况下,仍然能够得到较高的估计精度。“天巡一号”微小卫星在基本任务和探索性任务两个任务阶段中可分别得到普通遥感图像和侦察遥感图像。对于普通遥感图像,本文研究了预测编码+YEZW图像压缩方法,将图像经过小波变换后的小波系数的高频部分和低频部分分开来进行处理,低频系数部分由于包含了大部分的图像信息能量,因此采用几乎无损的预测编码方式,高频数据采用YEZW编码方式,在相当的压缩比条件下,提高了压缩图像的峰值信噪比。对于侦察遥感图像从最大限度地保护侦察对象为出发点,提出了一种基于侦察图像的压缩编码方法,将疑似目标图像和背景图像分离后,分别进行压缩编码,使得在不同的压缩比条件下,侦查目标图像信息都能够做到无损。针对目前卫星的故障诊断都是针对各个子系统进行的特点,设计了一种故障综合诊断的方法,为了减小诊断中的计算量,提出将故障的诊断分成两步完成,第一步实现故障的子系统级定位,第二步实现故障的部件级定位,使得故障的诊断更加准确。更多还原

【Abstract】 Not only in the development of single star but alse in the development of micro-satellite formation flying and constellation systems the demands on intelligence and autonomous in micro-satellite keeping operator are increasing. This paper just study on some key techniques of‘TX-1’micro-satellite keeping operator which is being researched in Nanjing University of Aeronautics and Astronartics.The mission of‘TX-1’micro-satellite is divided into nine tasks. The heuristic scheduling method of‘Merger-Threshold’is applicated to distribute these nine tasks. According to the shortcomings of the Merger-Threshold scheduling method a scheduling method based on genetic algorithm is studied and applicated to optimize the scheduling. In order to realize dynamic tasks scheduling, an intelligent autonomous scheduling method based on fuzzy neural network is presented in this paper. In this method the order requirement of some tasks and the different working conditions of some tasks are both consided. This method firstly classifies tasks according to the working order of tasks, then automatically decides the present reqirement of satellite by the present conditions and finally realizes dynamic tasks scheduling.There are many temperature measurement points in satellite, so the hardware redundant design method will increase much power consumption, volume and weight of satellite. Therefore, a redundant design method based on data fusion is presented in this paper. Temperature sensors are devided into several groups according to the positons. The sensor in one group offers reference to each other. Based on the references, the credibility of each sensor can be calculated. When the credibility of one sesor is too low, the measurement result of this sensor is ignored and the back up senson is opened. By this way, the reliability of temperature measurement is improved.In normal operation of the satellite, in order to induce communication data and simplify monitoring task, several sensors are needed to give the estimate value of the temperature area. In this paper a data fusion method based on arithmetic average and batch estimate and a data fusion method based on Bayesian Estimation are studied and applicated to improve the accuracy of temperature measurement. Especially when some fault in one sensor, the estimate value still can reach high accuracy.In the basic task stage of‘TX-1’micro-satellite ordinary remote sensing images will be achieved and in the exploratory task stage reconnaissance remote sensing images will be achieved. According to the ordinary remote sensing images, an improved EZW algorithm is studied. First, the image data are converted into lower frequency subimage and high frequency subimages by wavelet transform. Most of energy of images is in the low frequency subimag, so predicting coding method which almost loses no energy is used to compress low frequency subimage. And an improved EZW coding method is used to compress high frequercy subimages.According to the reconnaissance remote sensing images the most important element is to protect reconnaissance subimage. So an image compression method based on protecting reconnaissance subimage is presented. In this method, different methods are used to compress reconnaissance subimage and background subimage.Presently, most fault diagnosis system is carried out according to one subsystem of satellite. In this paper fault diagnosis method according to the whole satellite is presented. In order to decrease the amount of calculation the process of diagnosis is devided into two steps. In the first step fuzzy clustering method is used to estimate which subsystem works abnormally and in the second step the specific component or part with fault can be confirmed.更多还原