基于改进Faster R-CNN的失效卫星部件检测方法

基于光学图像对失效卫星部件的精确检测可以为失效卫星的定位与捕获等任务提供支撑。然而,失效卫星部件多为密集小目标,且其光照条件变化较大,这导致一般主干网络出现特征表征分辨率低、小目标漏检等问题。针对上述问题,提出了一种基于改进Faster R-CNN的失效卫星部件检测方法。首先,该方法在Faster R-CNN的基础上,融合高分辨网络构建新的特征提取主干网络,以获得可靠、高分辨率的特征表达式。其次,在模拟真实空间环境的条件下,利用1∶1的嫦娥卫星模型构建了一个信息丰富的失效卫星数据集。用该数据集进行验证,结果表明：该方法的平均精度为93.6%,与Faster R-CNN和Cascade R-CNN相比,对小部件检测的准确率与召回率分别平均提高了9.8%与5.4%。该方法可有效检测失效卫星部件。     

可靠性试验评定中相关专家信息融合

专家信息的相关性会对专家信息融合的结果产生影响.提出基于主成分分析的相关专家信息融合方法.首先给出可靠性试验评定中专家信息的提取融合流程,然后采用主成分分析方法对专家信息的相关系数矩阵进行分析,接着利用分析的结果对专家信息转化成的概率分布进行融合,最后给出了一个应用实例.     

Oblique shock train motion based on schlieren image processing

In this paper, shock train motion in a Mach number 2.7 duct is studied experimentally, and large numbers of schlieren images are obtained by a high-speed camera. An image processing method based on Maximum Correlation Detection (MCD) is proposed to detect shock train motion from the schlieren images, based on which the key structures, e.g., separation positions and separation shock angles on the top and bottom walls, can be analysed in detail. The oscillations of the shock train are generated by rhombus and ellipse shafts at various excitation frequencies. According to the analysis of MCD results, the distributions of the frequency components of shock train oscillation generated by the two shafts are distinctly different, in which the motion generated by the ellipse shaft is much smoother; shock train motion is mainly characterized by the oscillation of separation position while the separation shock strength is not so sensitive to downstream disturbance; there is a hysteresis loop relation between the downstream pressure and separation position.     

毫米波亚毫米波波导类组件制造工艺进展

综述了毫米波亚毫米波波导类组件制造工艺的国内外最新研究进展，将当前最主要采用的制造工艺进行了分类阐述，包括机械加工改进与数控超精密工艺、电化学工艺、微机电系统（micro-electro-mechanical system， MEMS）工艺、高能束流工艺、3D增材制造工艺。介绍了各分类工艺应用中最新进展的典型成果，讨论了其工艺方法的优势与局限性，并依据波导类产品的应用需求对其工艺研究与发展的趋势开展了分析。结合毫米波亚毫米波波导类产品的应用场景与相关产品未来的发展趋势，进一步提出了各分类工艺在该领域未来可能的发展方向。     

基于Arinc661的CDS标准部件库的设计与实现

本文提出了一种Arinc 661规范的解决方案,它采用VapsXT为工具,针对座舱显示应用,设计并实现了一个符合规范的座舱显示系统标准部件库框架,为研发航电显示产品提供了更为便利的平台.实验结果表明,通过本文提出的座舱显示系统内核标准部件库框架,可以方便地对部件库进行扩展.     

火箭基组合循环发动机引射过程准一维分析方法研究

目前对引射过程的分析模型还比较简单并且不准确。为提高引射分析模型精度，从引射过程的物理机制出发，以速度差异导致的质量交换作为一、二次流掺混过程的控制因素，以一次流动量覆盖整个法向截面时的参数剖面计算引射效果，建立了引射过程分析方法，并采用数值模拟和试验结果对方法进行了验证。结果表明：一、二次流掺混过程的计算模型能反映横截面上参数的变化趋势；当一次流处于过膨胀状态和轻度欠膨胀状态时，本方法计算的引射系数偏差在4.56%以内；当一次流处于严重欠膨胀状态时，在利用特征线法对截面静压进行校正以后，偏差在6%以内；以上精度均优于传统的Fabri模型。本方法还能准确地获得引射系统的临界背压，因此更适用于RBCC发动机这种有背压的应用场景。     

一种可靠性增长试验抽样数Max-Min准则的反馈修正方法

对成败型航空发动机部件可靠性增长试验抽样数的确定问题进行了研究。推导出了初始故障数服从典型离散分布时的最小化最大准则(Max-Min准则),提出了一种逐次抽样样本估计值反馈修正方法。最后给出了一个应用实例,验证了该方法的有效性。     

Contributions to data transmission at critical moments on unmanned aerial vehicles

As for unmanned aircraft, the knowledge of the aircraft performance is directly related with the navigation, guidance, and control system programming. Therefore, the measured data in each phase of the flight must be sufficiently precise to obtain a good characterization of aircraft. This article proposes new methods of sending information to ground, which make it possible to know the aircraft behavior accurately, and for this purpose, four contributions have been made for ALO (Avión Ligero de Observación, Spanish acronym for Light Observation Aircraft). Currently, the characterization is based on data obtained at ten samples per second, insufficient to acquire detailed knowledge of what happened during the whole flight of an aircraft. As a result of these contributions, many more samples per second of accelerations and angular velocities are obtained at the most critical moments of the flight, such as takeoff or landing. Among the improvements included are data compression techniques, providing references to locate the measured data in time and identifying labels of each parameter.