高性能计算和数据中心融合网络研究综述

随着高性能计算、大数据处理、云计算和人工智能计算呈融合发展趋势,高性能计算网络和数据中心网络的融合网络成为互连网络发展的重要趋势。分析当前融合网络研究现状;针对当前最具代表性的融合网络进行详细阐述,全面展示该领域的最新技术和动态;提出融合网络面临的技术挑战;基于技术挑战对融合网络的发展趋势进行展望,包括融合网络协议栈设计中融合与分化并存、基于在网计算实现融合网络性能加速、面向新兴应用需求优化融合网络性能。     

电磁脉冲对高频结构的热效应分析

对于Cerenkov型器件,由高频结构温升引发的热脱附会造成的射频击穿,进而导致输出功率降低和脉冲缩短。基于此,研究了电磁脉冲对高频结构的加热温升效应,推导了高频结构的温升公式,指出了理论公式的适用范围,并给出了数值求解试件温升的方法。以二周期1 MV·cm^-1的高频结构为例,进行了温升的数值计算和仿真研究。结果表明,脉宽100 ns的单脉冲对不锈钢试件的温升效应明显高于其他材料。当系统工作在高重频时,电磁脉冲的温升作用可能使金属材料达到气体脱附的阈值从而引发气体脱附形成局部高压。该研究可为高功率微波源中的气体脱附以及射频击穿等研究提供参考。     

一类非线性系统的自适应模糊反推近似滑模控制

针对一类含有未知非线性函数项和外界干扰的不确定纯反馈非线性系统,提出了一种自适应模糊反推近似滑模变结构控制方法。采用中值定理和隐函数定理使未知非仿射输入函数拥有显式的控制输入,利用模糊系统逼近未知非线性函数,动态面控制技术解决了反推设计中出现的"微分爆炸"问题。所提出的自适应近似滑模控制方案削弱了传统滑模控制中的抖振现象。从理论上证明了所设计的控制器能够保证闭环系统所有信号半全局一致终结有界。仿真算例验证了算法的有效性。     

基于滑模控制的某型起落架落震联合仿真

飞机起落架是飞机起落的安全保证。给出了在ADAMS/Aircraft中建立某型飞机起落架的虚拟样机模型,设计并实现基于磁流变液的飞机起落架滑模缓冲控制算法,通过MATLAB/Simulink和ADAMS完成了该型飞机起落架着陆缓冲联合仿真分析。经过联合仿真得到缓冲器的位移-时间曲线,速度-时间曲线,阻尼力-时间曲线等特性数据,通过与被动式起落架的缓冲性能进行比较,验证了设计的滑模控制算法能够有效提高飞机起落架着陆缓冲性能。     

“小机构”的“大智慧”——美国高能激光技术联合办公室

美国国防部在世纪伊始成立的高能激光技术联合办公室专职人员不到10人,却能高效调动创新体系中所有创新单元,长期保持战略先导地位。本文全面介绍联合办公室的使命职能、编制组成、运行方式、科研项目、代表性成果以及重要决策,旨在为技术创新和管理创新提供参考和借鉴。     

Sharp bounds for the reliability of systems and mixtures with ordered components

In this article, we study how to derive bounds for the reliability and the expected lifetime of a coherent system with heterogeneous ordered components. These bounds can be used to compare the performance of the systems obtained by permuting the components at a given system structure, that is, to study where we should place the different components at a system structure to get the most reliable system. Moreover, a similar procedure is applied to get bounds for mixtures and for the generalized proportional hazard rate model when the baseline populations are ordered. © 2017 Wiley Periodicals, Inc. Naval Research Logistics 64: 108–116, 2017     

On matched active redundancy allocation for coherent systems with statistically dependent component lifetimes

This study addresses the allocation of matched active redundancy components to coherent systems with base components having statistically dependent lifetimes. We consider base component lifetimes and redundancy component lifetimes which are both stochastic arrangement monotone with respect to a pair of components given the lifetimes of the other components. In this context, allocating a more reliable redundancy component to the weaker base component is shown to incur a stochastically larger system lifetime. Numerical examples are presented as an illustration of the theoretical results.     

Civilian Cooperation and Non-Cooperation with Non-State Armed Groups: The Centrality of Obedience and Resistance

Abstract

Terms like ‘support’ and ‘collaboration’ are often used interchangeably to denote a loose set of acts or attitudes that benefit non-state armed groups (NSAGs). However, these terms are seldom defined, and the alternatives available to civilians are rarely identified. Moreover, existing approaches overlook that the interaction between civilians and NSAGs is often one between ruler and ruled, which makes obedience and resistance central. This paper proposes to conceptualize the choices available to civilians as forms of cooperation and non-cooperation, offers a typology, and discusses the implications for theory building on civilian and NSAG behavior, and on the functioning of armed social orders.     

布拉格衍射型冷原子干涉重力仪关键实验条件研究

文章介绍了一种基于n阶布拉格衍射的新型冷原子干涉重力仪,可以进一步提高现有拉曼跃迁型原子干涉重力仪的测量灵敏度和稳定度。在介绍布拉格衍射型原子干涉重力仪基本原理的基础上,建立了原子平行驻波入射的时间型布拉格衍射冷原子干涉重力仪理论模型,分析了实验所需的关键条件,包括原子团纵向温度,布拉格激光直径、曲率半径、频率、强度以及时序等。与已有实验数据的对比结果表明,本文所建模型合理,所得结论能够为实际构造一台布拉格衍射型冷原子干涉重力仪提供有意义的指导。     

并行程序运行故障原因识别

高性能计算系统的复杂性和规模的不断增长使得系统的平均无故障时间越来越短,因此系统的硬软件故障导致并行程序运行出错的概率随之增加。此外,并行程序本身可能存在的编程错误也会导致运行出错。由于处理上述两类故障原因的措施迥异,所以在程序运行出现故障时,用户需要关注故障原因的类别。针对这一问题,设计和实现了一种基于作业管理系统Slurm的并行程序运行故障原因识别系统。通过对Slurm进行扩展,监控作业状态,重提交和重运行作业。根据作业运行结果,区分故障原因类别。故障注入方式进行的实验表明,该系统具有较高的识别准确率。