一种基于在轨深度学习的压缩率确定方法

如何对遥感图像中的重要目标进行精确的识别分类,是卫星遥感领域的一个难点和重要的研究方向。深度学习能较好解决识别分类的问题,但其学习模型有大量参数需要确定,会消耗大量计算和存储资源,不利于在轨实现。模型压缩是降低资源需求的有效方法,但会导致分类准确率降低。剪枝是模型压缩的主要方法之一,目前剪枝技术大多研究的是在降低计算量的情况下如何减少准确率损失,如何确定压缩率是有待研究的问题。本文提出了一种通过函数拟合准确率与压缩率关系的方法,可以据此确定相应的压缩率,并对不同的压缩方法进行比较。仿真结果表明:该方法的函数模型可以在不同场景下用较少的点拟合出准确率与压缩率关系曲线,且均方根误差最大为1.09,平均值为0.51,拟合效果较好,可据此针对不同的应用条件与需求确定相应的模型压缩率。     

Role of the inter-pass cooling rate in recrystallization behaviors of Ni-based superalloy during interrupted hot compression

The microstructural evolution of a Ni-based superalloy under interrupted hot compressive deformation with different cooling rates in the inter-pass stage is investigated. It is found that metadynamic recrystallization(MDRX) in the inter-pass stage is more sensitive to the accumulated strain than the deformation temperature which is above the recrystallization temperature. The variations of both the grain distribution and the texture intensity caused by MDRX during the interpass stage result in variations of the yield stress(YS) and the work hardening(WH) rate in each stage. Results also show that the MDRX process in the inter-pass stage has a considerable influence on the final microstructure of three-pass compression. The final grain distribution is more uniform,and the compression texture gradually transforms into recrystallization texture with an increasing degree of MDRX. In order to predict the MDRX fraction in the inter-pass cooling stage, a modified kinetic equation is established, which can reasonably predict the MDRX behavior under multi-pass compression with different conditions in the inter-pass stage. Meanwhile, the influence of the interpass cooling stage on the mechanism of dynamic recrystallization(DRX) is studied. It is universally acknowledged that the discontinuous dynamic recrystallization(DDRX) process is the major deformation mechanism for the Ni-based superalloy. However, the continuous dynamic recrystallization(CDRX) process is promoted in the compression stage with a decrease of the cooling rate in each inter-pass stage.     

Design and simulation of an innovative cylinder fabricated by selective laser melting

Additive Manufacturing (AM) processing has attracted wide attention due to its advantages, e.g., rapid prototyping and free design. In aeronautic and astronautic engineering, it is especially attractive to apply AM processing, especially Selective Laser Melting (SLM), into primary structures that always occupy most of the whole mass of aircraft or spacecraft. In this paper, an innovative lattice sandwich structure that is suitable to be manufactured with SLM is adopted to design the typical cylinder under axial compression. Firstly, the mechanical properties of pyramidal lattice are investigated, and the bearing capacity of the lattice sandwich cylinder is given analytically for each failure mode. Secondly, a composite cylinder is re-designed in the form of lattice sandwich structure, which sizes are ultimately obtained through optimization. Finally, finite element model of the innovative cylinder is established and several kinds of imperfection are introduced to simulate the defects caused by SLM. Numerical result shows that the lattice sandwich design can obviously enhance the bearing capacity of cylindrical shell even though the original one is designed with composite material. In other words, SLM has great potential in improving the primary structure of spacecraft in the future.     

飞机装载状态对机轮载荷和起降安全性影响的定量分析

飞机的装载状态直接决定了飞机的起落架压缩量和机轮载荷,同时起落架的压缩量还影响着飞机的起降稳定性。本文分析了机轮载荷的理论值偏离实测值的原因,利用解析几何方法对理论算法进行修正,开发了基于Visual Basic的飞机起落架压缩量与轮载分析程序,对各种可能的装载状态进行分析,得到了各状态的机轮修正载荷和起落架稳定性参数,计算结果与实测值吻合的较好。     

高超声速侧压式进气道溢流特性研究

首先介绍了高超声速侧压式进气道的构型, 定义了侧压式进气道可能存在的两种溢流模态———构型溢流和压差溢流;然后对构型溢流进行了理论分析和计算, 阐明了侧板后掠的侧压进气道设计参数对构型溢流影响;最后对不同侧板配置方式的侧压式进气道进行了数值模拟, 通过对比分析发现由唇口板位置所决定的溢流窗大小对进气道溢流特性的影响显著.      

基于小波变换的数字地图数据压缩技术研究

针对数字地图数据的特点 ,提出了一种基于小波变换的数字地图压缩方法。文中主要研究了小波滤波器的选取、标量量化、自适应算术编码、压缩文件存储格式等问题 ,并就在给定精度要求的前提下如何获得最大压缩比的问题进行了深入研究。文末应用该方法对 5种不同类型的数字地图数据进行了压缩与解压缩测试。结果表明 ,该方法与其他方法相比较 ,能够在相同失真的情况下产生更大的压缩比。     

干涉SAR原始数据压缩算法的比较研究

为了更合理地选择有利于星载合成孔径雷达干涉技术的原始数据压缩算法,文章仿真了三种合成孔径雷达原始数据压缩算法,并从信号量化噪声比、相对高程误差等角度分析比较了算法的性能。仿真结果表明三种算法各有其适用范围,块自适应量化(BAQ)仍是压缩比不高时的首选,块自适应矢量量化(BAVQ)在压缩比高时性能较好,而约束式极坐标量化(RPQ)的复杂度和性能均介于二者之间。     

变循环压缩系统的一体化通流设计方法

以双外涵和三外涵两款典型的涡扇发动机的变循环压缩系统为研究对象发展了一个基于流线曲率法的一体化通流设计方法,以期简化变循环压缩系统的设计流程,缩短设计周期.通过一个三外涵变循环压缩系统的设计实例,验证了方法的正确性和程序的功能完整性,得到了以下主要结论:①第1,2外涵道的连通使压缩系统各部件之间的气动关系变得复杂,正确处理各部件之间的匹配关系是一体化设计的关键;②流线曲率法适用于变循环压缩系统的通流设计,其流量边界条件可方便地用于模拟前可调面积涵道引射器的调节功能;③第1,2外涵道的掺混过程对上游叶片区域的流场影响微弱,采用简单的掺混模型能够满足计算需求.     

层状球面弹性轴承刚度设计、仿真与试验

从理论上推导了层状球面弹性轴承各橡胶层的压缩刚度和扭转刚度计算公式,通过具体算例,将层状球面弹性轴承各橡胶层压缩刚度和扭转刚度的理论计算结果与仿真结果进行了对比分析.结果表明:各橡胶层压缩刚度的理论计算结果与仿真结果基本吻合,最大误差为3.98%.各橡胶层扭转刚度的仿真结果与理论计算结果存在一定偏差,从小接头开始的前半部分橡胶层扭转刚度的仿真结果大于理论计算结果,最大误差为33.3%;后半部分橡胶层扭转刚度仿真结果小于理论计算结果,最大误差为32.8%.压缩刚度对压力变化表现的非线性特性不明显,扭转刚度则随扭转角的增大,其非线性特性变得越显著.但通过理论方法、有限元仿真方法得到的层状球面弹性轴承等效压缩刚度和等效扭转刚度与试验结果吻合良好.      

一种交织铺层结构层合板性能

结合复合材料自动纤维铺放(AFP)技术,提出了一种交织铺层结构层合板成型方法,制备了非交织、交织正交层合板、非交织、两向交织和四向交织准各向同性层合板复合材料,并对交织铺层结构复合材料的层间结构和性能进行了分析研究。结果表明:交织正交层合板的拉伸、压缩、弯曲性能较非交织板性能均有所下降,但其层间剪切性能有明显提高,提高幅度约为16%;随着交织铺层组厚度的增加,交织层合板的拉伸性能呈下降趋势;交织层合板的开孔后拉伸、开孔后压缩和冲击后压缩强度保持率均高于非交织板,且冲击后分层损伤面积明显低于非交织板。交织铺层结构层合板相对于普通非交织层合板具有更好的损伤容限能力。