基于迁移学习的小样本DGA恶意域名检测方法

域名生成算法（DGA）存在变化多、部分类别样本难获取的特点,使得采用传统机器学习的恶意域名检测模型准确性不高。提出一种基于迁移学习和多核CNN的小样本DGA恶意域名检测模型。该模型将目标域名映射到向量空间中,使用样本充足的DGA种类进行预训练,并迁移预训练得到的参数到小样本检测模型。采用多核CNN小样本分类模型根据发音习惯进行域名特征提取并分类。通过实验对比发现,无知识迁移的小样本分类模型只有11类域名准确率超过92%,经过迁移学习的多核CNN模型20类准确率超过92%,11类准确率超过97%,检测效果接近数据充足时的分类效果。     

GBAS与ILS联合导航系统数据融合方法与验证

电离层延迟和多径干扰分别是陆基增强系统（ground-based augmentation systems, GBAS）和仪表着陆系统（instrument landing systems, ILS）的主要干扰源。在实现GBAS与ILS时间对准和空间对准的基础上,分析了电离层风暴对GBAS以及多径干扰对ILS的影响;通过深入解析从模拟飞行软件X-Plane采集的多次进近着陆飞行数据,建立了飞机精密进近着陆阶段的运动模型;引入了一种基于改进的Sage-Husa自适应滤波和方差估计学习算法的GBAS与ILS数据融合方法;基于X-Plane及自行开发的数据采集模块构建了实验验证系统,并针对GBAS不受电离层风暴的影响、ILS受弱多径效应影响,GBAS受电离层风暴的影响、ILS受弱多径效应影响,以及GBAS不受电离层风暴的影响、ILS受多径效应影响三种情况展开仿真实验。实验结果表明,改进的Sage-Husa算法能有效抑制电离层风暴对GBAS的影响以及多径对ILS的干扰,使GBAS与ILS联合后的导航系统在相应干扰条件下,能为飞机提供满足I类精密进近精度要求的水平和垂直引导,有效提高精密进近运行的生存能力。      

An effective approach to reinforced closed-cell Al-alloy foams with multiwalled carbon nanotubes

Exploring the reinforcing role of carbon nanotubes to obtain materials (polymers, metals, ceramics) with enhanced properties has been often attempted, but the success is strongly limited by the dispersing degree of carbon nanotubes. Here we report on an innovative colloidal approach to disperse the carbon nanotubes in the powders mixture of the precursor materials in order to profit from their reinforcing potential and obtain a new class of closed-cell metal foams. The feasibility of the proposed approach was demonstrated for aluminium foams reinforced with multi-walled carbon nanotubes. These nanocomposite metal foams synergistically combine the remarkable properties of both metal foams and carbon nanotubes. The results indicate that the tubular structure of carbon nanotubes is preserved throughout the entire the process. The carbon nanotubes are individually dispersed, stretched and randomly aligned in the aluminium-matrix of these closed-cell foams, thus potentiating their homogeneous 3D reinforcing role. Accordingly, the Vickers micro-hardness of the closed-cell foams was greatly enhanced.     

激光选区熔化纳米SiC/AlSi7Mg复合材料微观组织及力学性能EI北大核心CSCD

采用机械混合法制备纳米SiC/AlSi7Mg混合粉末,利用激光选区熔化技术(selective laser melting,SLM)成形纳米SiC颗粒增强AlSi7Mg复合材料,观察和分析试样的相对密度、物相和微观组织,并测试材料的硬度和拉伸性能。结果表明:SLM成形纳米SiC/AlSi7Mg复合材料试样的相对密度随着扫描速度和扫描间距的增大均呈现先增加后减少的趋势,相对密度最高可达99.75%;试样微观组织与SLM成形铝合金相似,Si相呈网状结构均匀嵌入α-Al基体中,且在Al基体中存在与Si分布相似的纳米SiC团聚物及Mg_(2)Si相;与AlSi7Mg相比,复合材料微观组织由柱状晶转化为等轴晶,且晶粒明显细化(平均晶粒尺寸为1.36μm);由于SiC的加入,产生细晶强化和固溶强化,试样的硬度和强度均明显提高,硬度最高达到137.3HV,抗拉强度达到448.3 MPa,屈服强度达到334.7 MPa,但伸长率下降至3.9%,断裂模式主要为脆性断裂。     

2A12航空铝合金多轴疲劳试验及应力准则寿命预测模型研究

针对多轴疲劳失效问题,选取2A12航空铝合金进行应力幅比变量、相位差变量和平均应力变量的多轴疲劳试验。对常用的3种多轴疲劳应力准则寿命预测模型（即Lee准则、Carpinteri准则和Sines准则）进行讨论,并通过引入应力幅比参量和相位差参量,提出基于Carpinteri准则的修正模型。将不同条件下2A12航空铝合金的试验寿命与不同应力准则下模型的预测寿命进行比较,结果表明：Lee准则对上述多轴疲劳试验的预测结果过于危险;Carpinteri准则和Sines准则由于未考虑拉-扭应力幅比和相位差因素,预测寿命与实际寿命相比均出现了较大偏差;修正后的应力准则寿命预测模型在不同条件下90%的寿命预测数据在两倍误差带内。     

Selective deposition of diamond films on insulators by selective seeding with a double-layer mask

Polycrystalline diamond films have been patterned on Si₃N₄/Si and SiO₂/Si substrates by selective seeding with a double-layer mask via hot-filament chemical vapor deposition. High quality in the patterned diamond films and high selectivity were obtained by the process. The diamond films deposited on the insulators at different CH₄/H₂ concentrations were studied by scanning electron microscopy and Raman spectroscopy. The process proved to be far less damaging to the substrates, and yet effective in developing patterns of diamond films on a large and different substrate.     

基于有效信号提取的白光干涉信号快速处理方法

针对白光扫描干涉术在垂直大范围扫描过程中的测量效率问题,提出了一种基于有效信号提取的白光信号快速处理方法。使用白光LED光源,建立了双峰频谱分布模型,并使用该模型进行了仿真实验,给出了不同算法在特定采样步长下所需的最小采样区间长度确定方法,在不改变测量精度的条件下减少了解算所需的数据量,并使得轮廓解算可以在采样结束前开始。针对垂直大范围扫描过程中的背景值波动问题,给出了一种基于背景集合的白光干涉信号背景分离方法,并通过与几种常见信号处理方法的比较,证明了背景值提取的完整性,有效消除了背景值波动现象对测量精度造成的影响。最后将上述方法应用于自主设计开发的白光轮廓仪,使用傅里叶变换法精确测量了U盘接口表面形貌特征。测量结果表明:从扫描开始到获得表面高度信息的总时间减少了49.02%。     

A study of aluminium foam formation—kinetics and microstructure

Aluminium foams were produced by applying the powder compact melting method, i.e. by mixing metal powders and powdered gas-releasing blowing agents and pressing them to a foamable precursor material after this. The resulting precursor was then foamed by heating it up to above its melting point inside an “expandometer”, which allowed for the volume and temperature to be measured throughout the entire process. The present studies comprise the effects of the aluminium alloy composition (AlSi7 and 6061), some of the pressing parameters of the foamable precursor material, the foaming temperature and the heating rate during foaming on the expansion behaviour of the foam. Moreover, the morphological and microstructural evolution of metal foams is investigated.