Tomographic determination of the spatial distribution of water vapor using GPS observations

With the advent of the GPS navigation system, a promising ground based technique has been introduced which makes it possible to estimate the amount of water vapor in the troposphere from operational GPS networks at relatively low additional costs. While the estimation of the integrated amount is currently well established, the determination of the spatial water vapor distribution and its temporal variation are still a major challenge. To account for the vertical resolution, several tomographic approaches were pursued. We developed the software package AWATOS (atmospheric water vapor tomography software) which is based on the assimilation of double differenced GPS observations. Applying a least-squares inversion, the inhomogeneous spatial distribution of water vapor is determined. An extensive investigation has been carried out in Switzerland. GPS measurements are performed by the dense permanent Swiss national GPS network AGNES of the Swiss Federal Office of Topography (swisstopo). A total of 40 equally distributed water vapor profiles have been estimated on an hourly basis. For the purpose of validation, 22 radiosonde profiles were used at the GPS and meteorological station Payerne. Furthermore, data of the numerical weather model aLMo (alpine model in Switzerland, MeteoSwiss) were compared with the tomographic results. An overall good agreement of the three methods with an rms of better than 1.6 g/m³ absolute humidity was achieved. The results show that AGNES can be used as a dedicated network for the purpose of GPS-tomography, using a horizontal resolution of approximately 50 km and height layers of 300–500 m thickness in the lower troposphere.     

Improved orbit predictions using two-line elements

The density of orbital space debris constitutes an increasing environmental challenge. There are two ways to alleviate the problem: debris mitigation and debris removal. This paper addresses collision avoidance, a key aspect of debris mitigation. We describe a method that contributes to achieving a requisite increase in orbit prediction accuracy for objects in the publicly available two-line element (TLE) catalog. Batch least-squares differential correction is applied to the TLEs. Using a high-precision numerical propagator, we fit an orbit to state vectors derived from successive TLEs. We then propagate the fitted orbit further forward in time. These predictions are validated against precision ephemeris data derived from the international laser ranging service (ILRS) for several satellites, including objects in the congested sun-synchronous orbital region. The method leads to a predicted range error that increases at a typical rate of 100 m per day, approximately a 10-fold improvement over individual TLE’s propagated with their associated analytic propagator (SGP4). Corresponding improvements for debris trajectories could potentially provide conjunction analysis sufficiently accurate for an operationally viable collision avoidance system based on TLEs only.     

产品寿命预测

对基于失效或故障时间数据的产品寿命预测的基本概念、重要意义、方法以及与其它学科的关系等进行了系统评述.     

跳跃式返回航天器落点预报技术研究

建立了航天器再入动力学模型,提出了使用动能定理计算再入航天器质阻比和升阻比的方法。针对着陆场系统无法获取返回舱升力控制模型的实际情况,提出了一种使用理论机动弹道进行修正的再入航天器航迹预报算法;使用“嫦娥五号”再入段实测数据进行验证,结果显示本文算法显著提高了跳跃式返回航天器落点预报精度。     

基于排队论的终端区航班进场时间预测

为了提高终端区航班运行效率,本文将基于排队模型对终端区航班进场时间进行预测,以便尽早发现进场时间过长、进场效率低的航班,从而提前采取应对措施,确保航班正常运转。以天津滨海国际机场为例的研究结果表明,终端区内航班到达规律服从泊松分布,服务时间服从Gamma分布,本文建立的一般服务时间M/G/1排队模型能较好地预测航班进场时间,以15 min为单位的终端区航班进场时间预测的平均绝对误差在1 min左右。本文所建排队模型通过抓住进场交通流的整体流动特征来对进场时间进行预测,降低了现有模型难度,预测结果可以为改善管制工作带来更直接的指导建议。     

基于时间序列数据挖掘的旋转机械故障预报

提出一种基于时间序列数据挖掘的故障预报新方法。把故障前兆因子作为一种暂态,根据旋转机械轴承振动的实验数据建立时间序列．利用时延嵌入的方法重构状态空间,在状态空间中使用遗传算法搜寻最优暂态束．组成暂态集。用暂态集对旋转机械轴承振动的测试数据进行分析．判断是否为故障前兆因子．从而实现故障预报。     

神经网络BP算法预测可靠性增长

叙述可靠性增长的基本概念和神经网络BP算法,着重探讨BP算法用于可靠性增长预测的可能性。通过两个实例的预测结果,与常用的可靠性增长参数模型Duane,Gompertz的结果相比较。实例表明,该方法是可行的,而且是有效的,适应性强。可以预料,它不但能用于仪器设备试制过程中预测可靠性增长,而且可用于武器装备在储存过程之中可靠性下降的预估。     

应变疲劳等损伤曲线及其应用

利用疲劳损伤的剩余寿命定义，构造了应变疲劳等损伤曲线，并给出了其近数学描述，通过等损伤曲线，研究了应变疲劳累积损伤规律，建立了应变疲劳累积损伤公式。利用该公式来估算变幅应变疲劳寿命，仅需用材料的ε－Ｎ曲线，十分方便；其有效性得到２０２４－Ｔ４铝合金分级加载应变疲劳试验数据的验证。     

一种基于模糊聚类的组合BP神经网络数据挖掘方法

介绍了一种基于模糊聚类的组合BP神经网络的数据挖掘方法，并给出了该方法的模型和启发式BP改进算法Heuristicbp。且将其应用于数学函数值预测中，取得了学习时间短和预测精度高的效果，证明该方法是有效的，具有较高的实际应用性。