基于K Framework的向量化机器学习指令语义形式化

ARM针对ARMv8.1-M微处理器架构推出基于M-Profile向量化扩展方案的技术,并命名为ARM Helium,声明能为ARM Cortex-M处理器提升达15倍的机器学习性能.随着物联网的高速发展,微处理器指令执行正确性尤为重要.指令集的官方手册作为芯片模拟程序,片上应用程序开发的依据,是程序正确性基本保障.主要介绍利用可执行语义框架K Framework对ARMv8.1-M官方参考手册中向量化机器学习指令的语义正确性研究.基于ARMv8.1-M的官方参考手册自动提取指令集中描述向量化机器学习指令执行过程的伪代码,并将其转换为形式化语义转换规则.通过K Framework提供的可执行框架利用测试用例,验证机器学习指令算数运算执行的正确性.     

DNS of spark ignition and edge flame propagation in turbulent droplet-laden mixing layers

A parametric study of forced ignition at the mixing layer between air and air carrying fine monosized fuel droplets is done through one-step chemistry direct numerical simulations to determine the influence of the size and volatility of the droplets, the spark location, the droplet-air mixing layer initial thickness and the turbulence intensity on the ignition success and the subsequent flame propagation. The propagation is analyzed in terms of edge flame displacement speed, which has not been studied before for turbulent edge spray flames. Spark ignition successfully resulted in a tribrachial flame if enough fuel vapour was available at the spark location, which occurred when the local droplet number density was high. Ignition was achieved even when the spark was offset from the spray, on the air side, due to the diffusion of heat from the spark, provided droplets evaporated rapidly. Large kernels were obtained by sparking close to the spray, since fuel was more readily available. At long times after the spark, for all flames studied, the probability density function of the displacement speed was wide, with a mean value in the range 0.55-0.75S_L, with S_L the laminar burning velocity of a stoichiometric gaseous premixed flame. This value is close to the mean displacement speed in turbulent edge flames with gaseous fuel. The displacement speed was negatively correlated with curvature. The detrimental effect of curvature was attenuated with a large initial kernel and by increasing the thickness of the mixing layer. The mixing layer was thicker when evaporation was slow and the turbulence intensity higher. However, high turbulence intensity also distorted the kernel which could lead to high values of curvature. The edge flame reaction component increased when the maximum temperature coincided with the stoichiometric contour. The results are consistent with the limited available experimental evidence and provide insights into the processes associated with ignition of practical spray flames.     

408XL系统V8.1软件升级

V8.1软件是法国SERCEL公司为408XL系统最新推出的软件。文章对408XL系统V8.1软件升级准备、升级步骤进行详细地阐述,提出升级过程中应注意的问题。同时对系统软件升级后遇到的技术问题,给出了解决方法和建议。     

油吸收脱甲烷系统的工艺模拟计算

使用化工流程模拟计算软件PR0Ⅱ8．1对先进的低投资乙烯技术（ALCET）中油吸收脱甲烷系统进行了模拟计算,结果表明,该流程满足工艺流程的要求。分析了油吸收脱甲烷系统相对于传统深冷分离工艺的优点,并对油吸收脱甲烷塔的吸收效果进行了实验考察。     

PowerBuilder8.1中EAServer组件的研究及实现

在详细阐述PowerBuilder8．1的组件技术的基础上，以高校社科科研管理系统中的查询模块为例，说明了如何在PowerBuider8．1和EAServer4．1．1的开发环境下创建组件，在服务器端部署组件，并在客户端调用该组件．     

PowerBuilder8.1中EAServer组件的研究及实现

阐述了PowerBuilder8．1的组件技术，以高校社科科研管理系统中的查询模块为例，说明了如何在PowerBuider8．1和EAServer4．1．1的开发环境下创建组件，在服务器端部署组件，并在客户端调用该组件。     

青藏高原北部地区M≥7强震有序网络结构及其预测研究

青藏高原北部地区是中国西部主要地震区,自1700年以来,该区M≥7强震具有显著的有序性,其主要有序值为106~107 a、77~78 a、53~54 a、26~27 a、10~11 a与3~4 a等.以翁文波信息预测理论为指导,将有序性分析与复杂网络技术相结合,努力探索具有中国特色自主创新的强震中长期预测方法,构建了青藏高原北部地区M≥7强震信息有序网络结构.在总结该区21世纪以来3次大震(2001年昆仑山8.1级大震、2008年汶川8.0级大震和2010年玉树7.3级强震)预测研究的基础上,根据所建强震有序网络结构提出新的预测意见:2014—2015年、2026—2027年和2030年前后该区仍有可能发生M≥7强震.研究结果表明此方法对于强震的中长期预测具有独特的效果.