Interface microstructure and formation mechanism of ultrasonic spot welding for Al–Ti dissimilar metals

The present study focuses on interface microstructure and joint formation. AA6061 aluminum alloy (Al) and commercial pure ti-tanium (Ti) joints were welded by ultrasonic spot welding (USW). The welding energy was 1100–3200 J. The Al–Ti joint appearance and in-terface microstructure were observed mainly via optical microscopy and field emission scanning electron microscopy. Results indicated that a good joint can be achieved only with proper welding energy of 2150 J. No significant intermetallic compound (IMC) was found under all con-ditions. The high energy barriers of Al–Ti and difficulties in diffusion were the main reasons for the absence of IMC according to kinetic ana-lysis. The heat input is crucial for the material plastic flow and bonding area, which plays an important role in the joint formation.     

注意力机制下的EMD-GRU短期电力负荷预测

为进一步提高短期电力负荷预测精度,构建一种基于注意力机制的经验模态分解(EMD)和门控循环单元(GRU)混合模型,对时间序列的短期负荷进行预测.首先,对负荷序列进行EMD,将数据重构成多个分量;再通过GRU提取各分量中时序数据的潜藏特征;经注意力机制突出关键特征后,分别对各分量进行预测;最后,将各分量的预测结果叠加,得到最终预测值.仿真结果表明:相对于BP网络模型、支持向量机(SVR)模型、GRU网络模型和EMD-GRU模型,基于EMD-GRU-Attention的混合预测模型能取得更高的预测精度,有效地提高短期电力负荷预测精度.     

基于Transformer神经网络模型的网络入侵检测方法

网络入侵检测一直以来都是网络安全中亟待解决的关键任务之一,传统网络入侵检测方法主要通过提取多维特征,采用机器学习方法构建检测模型,大多忽略了入侵行为的时间相关性.通过提取网络入侵行为的时序特征,设计基于降维特征的多头自注意力机制Transformer网络模型,以解决传统串行化时序神经网络模型不易收敛且时间开销较大的问题,通过选取最优的损失函数和训练参数进行并行化训练,实现网络入侵行为检测.实验结果表明,基于Transformer网络模型的网络入侵检测方法在多个数据集上均获得了99％以上的精度和检出率.     

一种含放大机构隔振器的动力学特性研究

利用四端铰接杆,提出一种含放大机构的新型隔振器.采用谐波平衡法获得隔振系统动态响应的解析表达式,并分别给出隔振系统的幅频响应、相频响应、绝对位移传递率、等效刚度系数和等效阻尼系数.理论计算结果与常规线性隔振器进行对比,并开展了验证试验.分析并给出系统参数对含放大机构隔振系统动态性能的影响规律.结果表明:与常规线性隔振器相比,在时域范围多频激励条件下含放大机构隔振系统惯性质量的绝对位移传递率幅值得到明显衰减;同样,在频域范围惯性质量的绝对位移传递率幅值也得到显著衰减,且谐振频率略微向高频移动;隔振系统设计参数对隔振性能均有较大影响,通过灵活调整隔振系统各设计参数可以实现不同的隔振目标.      

Effect of superplastic deformation on precipitation behavior of sigma phase in 3207 duplex stainless steel

Owing to excellent strength, toughness and corrosion resistance, duplex stainless steels(DSS) are widely used in constructional and petrochemical applications. Sigma phase, which has detrimental impact on the properties, is readily precipitated during hot working of DSS. However, precipitation behavior of sigma phase during superplastic deformation, which is the most significant processing method of DSS, is still unclear. In the current study,the effect of superplastic deformation on the precipitation behavior of sigma phase was investigated in 3207 duplex stainless steel. The result shows that superplastic deformation could prevent sigma phase precipitation generally by increasing mobility of grain boundaries and decreasing misorientation of the sigma phase boundaries, resulting in some sigma phase precipitated on the twin boundaries. Most of the sigma phase precipitated on ferrite-austenite interface with misorientation of 20–25°, while it precipitated in ferrite or austenite with the misorientation of 40°–45°. The orientation relationship between sigma phase and matrix matched well in austenite and on the ferrite/austenite interfaces, while it showed a small misfit in ferrite. The prevention effect of the superplastic deformation on the sigma phase precipitation was beneficial to quasi stable deformation stage,resulting in longer elongation.     

多孔材料的孔结构对其力学性能及破裂机制的影响

针对多孔材料的孔结构对其宏观力学性能及其破裂机制与耐久性的影响,对两种不同孔结构的多孔材料进行了单轴压缩和冻融循环试验,建立了能反映材料内部孔隙结构的三维非均匀数值模型,对两种多孔材料在单轴压缩下的破裂机制进行了分析.实验结果表明:孔结构对于金尾矿多孔材料的力学性能有显著影响.较大的孔隙率除会显著降低材料的抗压强度外,吸能性能会显著增加;在±25℃范围内冻融循环10次条件下的试验结果表明,当金尾矿多孔材料的孔隙率较小且为闭孔结构时,孔洞因冻融造成的孔壁颗粒脱落填塞而造成材料的抗冻性能有所提高;孔隙率较小时,金尾矿多孔材料在单轴压缩下的裂纹扩展主要从试件两端向试件中部扩展贯通为一条连续主拉裂纹,次生裂纹较少;而孔隙率较大时,压缩过程会出现大量微裂纹,最终主裂纹附近会伴随大量次生裂纹.     

宋式二级燕尾榫特殊结构的力学性能分析

为了更好地分析燕尾榫结构的力学性能,建立了二折线多参数力学结构,并进行了适用于木结构的力学分析试验.然后,根据力学平衡与变形协调关系公式,提出了燕尾榫节点计算公式.研究结果表明,燕尾榫的变形主要集中在榫头与卯口的接触处,当单向载荷变大时,其弯矩也变大,容易对接触面造成损伤.荷载-位移曲线和极限荷载-竖向荷载曲线显示,当竖向荷载增加时,极限荷载也随之增加;当极限载荷达到屈服临界点时,位移的增加量变小并趋于平稳.     

Enhanced resistive switching properties of HfAlOx/ZrO2- based RRAM devices

Resistive Random-Access Memory (RRAM) devices are recognized as potential candidates for next-generation memory devices, due to their smallest cell size, high write/erase speed, and endurance. Particularly, the resistive switching (RS) characteristics in oxide materials have offered new opportunities for developing CMOS-compatible high-density RRAM devices. In this study, the RS behavior of HfAlO_x/ZrO₂ thin films sandwiched structure between TiN bottom electrode and Au top electrodes were investigated. It was found that Au/HfAlO_x/ZrO₂/TiN stacks were superior in terms of RS performance when compare to Au/HfAlO_x/TiN memory stacks. The devices demonstrated a good resistance ratio of high resistance state and low resistance state ～10³ for Au/HfAlO_x/TiN and ～10⁵ for Au/HfAlO_x/ZrO₂/TiN stacks, respectively. Both stacks showed good retention characteristics (>10⁴ ?s) and endurance (>10³ cycles). The experimental current-voltage characteristics fitted with different conducting mechanisms, the linear lower bias region is dominated by ohmic conductivity, whereas the non-linear higher bias region was dominated by space-charge limited current conduction mechanism.     

一种带有灾难和伯努利机制的 M/M/1 队列

针对带有灾难和伯努利机制的模型在实际生活中的应用问题,提出了一种伯努利机制下具有灾 难、延迟维修、反馈和休假的单工作台队列。 当工作台运行时,灾难才会影响系统,此时,系统需要被维修,在场的所有顾客从系统中永远离开;工作人员对顾客完成一次服务后,可以选择休假或者继续服务;而接受这次服务的顾客,离开系统或者回到队首等待下次服务。 利用马氏链方法,对稳态下系统进行分析,得到平衡方程;对平衡方程求解,导出稳态下队列中顾客人数的 PGF,工作人员分别处于休假期、忙期、延迟期、维修期和空闲期的概率;根据强马尔可夫性求出稳态下逗留时间的分布;最后利用数值实验解释一些参数对系统中平均顾客人数的影响,验证了模型与方法的正确性。     

多注意力结合光流的视频超分辨方法

针对现如今监控摄像、卫星遥感以及视频娱乐等领域对视频图像的清晰度要求越来越高,而目 前大部分视频超分辨方法存在参数量大、恢复的视频存在抖动等问题,提出了一种多注意力结合光流的视频 超分辨方法,通过引入多个注意力包括空间注意力、通道注意力以及自注意力来提升超分辨性能。 具体而 言,作为一种特征加权的增强方法,这些注意力可以捕获视频帧的时空特征并增强自适应性和通道间的依赖 性,实现全局学习的功能;同时,提出双阶段特征对齐思路,首先利用光流对视频进行估计,进行第一阶段的 特征对齐,然后引入长短是记忆网络结构增强位置和通道的特征融合,进行第二阶段的特征对齐,以防止恢 复的视频帧出现抖动。 实验结果表明:该方法在评估标准和可视化效果方面都取得了令人满意的效果。