编排问题的一种方差最优算法——“手拉手”算法

介绍了编排问题中方差最优的均衡算法。此算法的特点是：通过量化工作得到完全规律编排工作的数学模型，求出模型中冲突段内每项工作最优调整位置，从而实现冲突调整后工作的目标位置与完全规律编排的位置之间的方差最小。     

Identification Method for RLG Random Errors Based on Allan Variance and Equivalent Theorem

An identification method using Allan variance and equivalent theorem is proposed to identify non-stationary sensor errors mixed out of different simple noises. This method firstly derives the discrete Allan variances of all component noises inherent in noise sources in terms of their different equations; then the variances are used to estimate the parameters of all component noise models; finally, the original errors are represented by the sum of the non-stationary component noise model and the equivalent model mixed out of the stationary and critically stationary component noises. Results of two examples for identification confirm the superiority of this approach regardless of the errors being stationary or not. The comparison of results of real ring laser gyro (RLG) errors processed by various methods shows that the proposed approach is more suited to depict the original noises than common ones.     

模糊随机可靠性灵敏度分析方法

 依据模糊随机可靠性灵敏度的基本概念,提出了模糊随机可靠性灵敏度分析的通用数字模拟法,推导了模糊随机可靠性灵敏度数字模拟解的方差和变异系数的计算公式。当模糊变量的隶属函数为正态型时,可以将模糊随机可靠性灵敏度转化为随机可靠性灵敏度,并利用复合函数求导法则求解模糊随机失效概率对正态型隶属函数分布参数的灵敏度。对于工程中常用的对称三角型隶属函数,提出了两种近似等价正态化变换基础上的模糊随机可靠性灵敏度分析方法,这两种将对称三角型隶属函数近似等价为正态型的方法分别是“3σ规则”法和“最大最小”法。算例结果表明：由于“最大最小”法比“3σ规则”法所得的等价正态隶属函数能在形状上更好地近似对称三角型隶属函数,因而“最大最小”变换法更适用于对称三角型隶属函数情况下模糊随机可靠性灵敏度的近似计算。     

相关噪声系统的卡尔曼滤波

利用Hilbert 空间的几何理论求出了噪声相关系统的卡尔曼滤波公式.在线性最小方差准则下,解决了状态噪声一步相关,相同时刻的测量噪声和状态噪声相关,以及测量噪声和状态噪声在相邻时刻(无论测量噪声时刻在前还是状态噪声时刻在前)均相关的情况下的滤波问题,并给出了递推滤波公式.     

The simulation of lunar gravity field recovery from D-VLBI of Chang’E-1 and SELENE lunar orbiters

The lunar gravity field is a foundation to study the lunar interior structure, and to recover the evolution history of the Moon. It is still an open and key topic for lunar science. For above mentioned reasons, it becomes one of the important scientific objectives of recent lunar missions, such as KAGUYA (SELENE) the Japanese lunar mission and Chang’E-1, the Chinese lunar mission. The Chang’E-1 and the SELENE were successfully launched in 2007. It is estimated that these two missions can fly around the Moon longer than 6 months simultaneously. In these two missions, the Chinese new VLBI (Very Long Baseline Interferometry) network will be applied for precise orbit determination (POD) by using a differential VLBI (D-VLBI) method during the mission period. The same-beam D-VLBI technique will contribute to recover the lunar gravity field together with other conventional observables, i.e. R&RR (Range and Range Rate) and multi-way Doppler. Taking VLBI tracking conditions into consideration and using the GEODYNII/SOVLE software of GSFC/NASA/USA [8 and 10], we simulated the lunar gravity field recovering ability with and without D-VLBI between the Chang’E-1 and SELENE main satellite. The cases of overlapped flying and tracking period of 30 days, 60 days and 90 days have been analyzed, respectively. The results show that D-VLBI tracking between two lunar satellites can improve the gravity field recovery remarkably. The results and methods introduced in this paper will benefit the actual missions.     

基于非下采样轮廓波的多源图像融合

利用非下采样轮廓波（NSCT）对源图像进行多尺度、多方向分解,对分解后的低频子带系数采用区域能量的融合规则,对高频子带系数采用基于领域方差加权平均的融合规则,对融合后的系数经过NSCT逆变换得到融合图像。实验结果表明,在主观视觉和客观评价指标上均优于其他对比方法。     

不连续观测数据对计算的原子钟稳定度影响研究

为评估原子钟稳定度,通常采用连续的、均匀的采样数据计算阿伦方差或哈达玛方差。在实际工程计算中,由于观测粗差剔除、测量仪器中断、卫星不可见等原因,经常导致原子钟的观测数据在时间上不连续,给原子钟稳定度评估带来影响。文中比较了不连续观测数据的几种常用处理方法,包括线性插值补齐、B样条插值补齐、不插值直接计算哈达玛方差等方法,分析了各处理方法对阿伦方差和哈达玛方差的计算影响,研究中断数据长度与稳定度评估的影响关系。采用的氢钟观测数据,对不连续观测数据的影响和处理效果进行统计和分析,将稳定度计算结果、计算精度损失比与理论分析进行对照和检验,得出了有益的结论。     

行星齿轮传动系统的瞬态热分析

行星齿轮传动系统在失去润滑后 ,将转化为一个具有热时变特征的系统 ,则可利用瞬态热分析法来求解该系统任意时刻的温度场分布。在求解过程中 ,基于的基本原理是节点的净热流量等于节点相关体积内能的增加 ;并运用数值方法 ,将时间 τ按一定步长离散成一时间序列 ,从而求得每个时刻的温度场 ,以预测该系统的生存能力及危险部位     

适用于全包线的航空发动机BP网络模型的动态辨识

为克服传统的发动机动态模糊辨识中存在的辨识精度低，辨识模型应用范围窄等不足，把对非线性系统具有高度逼近能力的神经网络应用于航空发动机动态特性的辨识，从而为发动机动态辨识开辟更为广阔的道路，采用均方差归一法的处理方法和BP算法的改进算法－输出端动量BP地，以某型发动机在飞行包线内某一飞行条件下的数据作为学习样本，辨识了发同的神经网络模型，在全包线范围内对该模型进行检验，结果表明，所得的发动机动态模型在全包线内都有很高的逼近精度，而且对噪声有很强的抑制能力。