基于1D CNN与XGBoost的恶意代码纹理检测

恶意代码数量已经呈现爆炸式增长,对于恶意代码的检测防护显得尤为重要.近几年,基于深度学习的恶意代码检测方法开始出现,基于此,提出一种新的检测方法,将恶意代码二进制文件转化为十进制数组,并利用一维卷积神经网络(1 Dimention Convolutional Neural Networks,1D CNN)对数组进行分类和识别.针对代码家族之间数量不平衡的现象,该算法选择在分类预测上表现良好的XGBoost,并对Vision Research Lab中的25个不同恶意软件家族的9458个恶意软件样本进行了实验.实验结果表明,所提的方法分类预测精度达到了97％.     

一种用于水声材料设计的动态参数测试方法

针对提高水声材料设计中动态参数输入精度的问题，提出了一种复杨氏模量及泊松比准确测试方法。对于复杨氏模量测试，通过将Williams-Landel-Ferry(WLF)方程引入到Havriliak-Negami(H-N)模型中，采用信赖域反射算法对未知参数进行拟合得到材料参数宽频域主曲线。对于泊松比测试，根据同一材料不同形状因子表观杨氏模量之比与泊松比存在唯一量化关系的特性，仅通过两种不同形状因子试样的准静态有限元模拟，获得表观杨氏模量比值与泊松比量化曲线。因此，根据橡胶样品表观杨氏模量测试结果，可以直接利用局部加权回归获得其泊松比。最后，将前述材料制成直径为55 mm、厚度为50 mm的声管样品，放置在水声管中进行吸声系数测试。同时，把橡胶的复杨氏模量和泊松比的测量结果输入到水-橡胶-水分层介质模型中进行吸声系数计算。结果表明两者吻合，验证了上述测试方法的正确性和有效性。     

The effect of alkali metal oxide on the properties of borosilicate fireproof glass: Structure,thermal properties,viscosity, chemical stability

The purpose of the current work was to research the effect of alkali metal oxide on the structure, thermal properties, viscosity and chemical stability in the glass system (R₂O–CaO–B₂O₃–SiO₂) systematically. Because the glass would emulsify when Li₂O was added to the glass batch, this article did not discuss Li₂O. The results showed that when the amount of Na₂O was less than 4 mol.%, there was a higher interconnectivity of borate and silicate sub-networks in glass, as more mixed Si–O–B bonds were present in glass. The glass samples exhibited excellent thermal properties and chemical stabilities. As the amount of Na₂O exceeded 4 mol.%, the interconnectivity of borate and silicate sub-networks was weakened. The thermal properties and chemical stabilities of the glass samples were reduced. The connectivity of the silicate sub-network was weakened slightly as the Na/K ratio varied, and the coefficient of thermal expansion (CTE) of the glass samples gradually increased, and the resistance to thermal shock (RTS) value gradually decreased. Moreover, the viscosity of the glass samples decreased with the ratio of Na/Si and Na/K increased.     

Enhanced electrochemical performance of LiNi0.8Co0.1Mn0.1O2 via titanium and boron co-doping

Titanium and boron are simultaneously introduced into LiNi_0.8Co_0.1Mn_0.1O₂ to improve the structural stability and electrochemical performance of the material. X-ray diffraction studies reveal that Ti⁴⁺ ion replaces Li⁺ ion and reduces the cation mixing; B³⁺ ion enters the tetrahedron of the transition metal layers and enlarges the distance of the [LiO₆] layers. The co-doped sample has spherical secondary particles with elongated and enlarged primary particles, in which Ti and B elements distribute uniformly. Electrochemical studies reveal the co-doped sample has improved rate performance (183.1 mAh·g^-1 at 1 C and 155.5 mAh·g^-1 at 10 C) and cycle stability (capacity retention of 94.7% after 100 cycles at 1 C). EIS and CV disclose that Ti and B co-doping reduces charge transfer impedance and suppresses phase change of LiNi_0.8Co_0.1Mn_0.1O_2.     

基于有限元的区域分解方法在永磁聚焦系统仿真中的应用

随着计算机技术以及并行求解技术的发展,区域分解方法越来越多地应用于计算电磁学的各个领域.针对微波管中的永磁聚焦系统仿真,该文提出一种基于有限元的非重叠区域分解方法,其引入一种新型传输条件,并采用内罚的方式推导出有限元弱形式.该区域分解法的最大优势是不需要引入多余的未知量,并且最终集成的有限元矩阵满足对称正定性,适合采用预处理共轭梯度法进行矩阵方程的求解.该文仿真了多个微波管永磁聚焦系统,并与商业软件Maxwell进行了详细的对比,结果表明所提出的区域分解方法和Maxwell精度相当,却拥有着更加优越的计算性能.