5G核心网业务模型的智能化预测研究

摘 要：核心网业务模型的建立是5G网络容量规划和网络建设的基础，通过现有方法得到的理论业务模型是静态不可变的且与实际网络存在偏离。为了克服现有5G核心网业务模型与现网模型适配性较差以及规划设备无法满足用户实际业务需求的问题，提出了一种长短期记忆（long short-term memory，LSTM）网络与卷积LSTM （convolution LSTM，ConvLSTM）网络双通道融合的 5G 核心网业务模型预测方法。该方法基于人工智能（artificial intelligence，AI）技术以实现高质量的核心网业务模型的智能预测，形成数据反馈闭环，实现网络自优化调整，助力网络智能化建设。     

Tuning of shear thickening behavior and elastic strength of polyvinylidene fluoride via doping of ZnO-graphene

ZnO rice like nonarchitects are grafted on the graphene carbon core via a rapid microwave synthesis route. The prepared grafted systems are characterized via XRD, SEM, RAMAN, and XPS to examined the structural and morphological parameters. Zinc oxide grafted graphene sheets (ZnO-G) are further doped in β-phase of polyvinylidene fluoride (PVDF) to prepare the polymer nanocomposites (PNCs) via mixed solvent approach (THF/DMF). β-phase confirmation of PVDF PNCs is done by FTIR studies. It is observed that ZnO-G filler enhances the β-phase content in the PNCs. Non-doped PVDF and PNCs are further studied for rheological behavior under the shear rate of 1–100 s⁻¹. Doping of ZnO-G dopant to the PVDF matrix changes its discontinuous shear thickening (DST) behavior to continues shear thickening behavior (CST). Hydrocluster formation and their interaction with the dopant could be the reason for this striking DST to CST behavioral change. Strain amplitude sweep (10⁻³% -10%) oscillatory test reveals that the PNCs shows extended linear viscoelastic region with high elastic modulus and lower viscous modulus. Effective shear thickening behavior and strong elastic strength of these PNCs present their candidature for various fields including mechanical and soft body armor applications.     

The influence of FeNi nanoparticles on the microstructures and soft magnetic properties of FeSi soft magnetic composites

In this work, a new type of FeSi/FeNi soft magnetic powder core (SMPC) was successfully fabricated by coating FeNi nanoparticles on the surface of FeSi micrometer powder. The effects of different contents of FeNi nanoparticles on the micromorphology, internal structures, and soft magnetic properties of SMPCs were studied. The results show that FeNi nanoparticles adhere to the surface of FeSi powder, which can effectively fill the air gap between FeSi powder and is beneficial to the compaction of the powder cores during the pressing process. Thus, the density of the SMPCs is increased. Compared to FeSi SMPCs, the comprehensive soft magnetic properties of FeSi/FeNi SMPCs have been greatly improved. When adding 15 wt% FeNi nanoparticles, the SMPCs exhibit excellent magnetic properties with high effective permeability (increased by 43.8 %) and low core loss (decreased by 22.1 %). The high performance FeSi/FeNi SMPCs prepared in this work are expected to be widely used in power choke coils, uninterruptible power supplies, and boosts and inverter inductors.     

Picosecond laser trimming of ceramic cores with porous multi-scale particle microstructure

The ceramic core, produced by hot injection molding, is one of the critical components for manufacturing high-performance aircraft engine turbine blades. However, the injection molding process will cause defects such as burrs and flashes in the fine structure of the formed ceramic core. Manual trimming is necessary, but the trimming quality is poor, and the yield is low. In this paper, the online trimming method of ceramic cores is studied. Based on the orthogonal experiment method, the optimal laser parameters for processing the ceramic core's porous multi-scale particle structure material were obtained. Further, the problems of the match head and tail phenomenon and dimensional accuracy improvement in trimming ceramic cores have been studied. A path optimisation method is proposed to improve the quality and accuracy of the trimming profile effectively. Finally, the overall process flow of ceramic core trimming is elaborated, and experimental verification is given. The results show that the ceramic core online trimming method proposed in this paper has advantages of high precision and high yield compared with the manual method, which will have substantial potential application value in the aviation field.     

Novel Bi-Doped Amorphous SnOx Nanoshells for Efficient Electrochemical CO2 Reduction into Formate at Low Overpotentials

Engineering novel Sn-based bimetallic materials could provide intriguing catalytic properties to boost the electrochemical CO₂ reduction. Herein, the first synthesis of homogeneous Sn_1−xBi_x alloy nanoparticles (x up to 0.20) with native Bi-doped amorphous SnO_x shells for efficient CO₂ reduction is reported. The Bi-SnO_x nanoshells boost the production of formate with high Faradaic efficiencies (>90%) over a wide potential window (−0.67 to −0.92 V vs RHE) with low overpotentials, outperforming current tin oxide catalysts. The state-of-the-art Bi-SnO_x nanoshells derived from Sn_0.80Bi_0.20 alloy nanoparticles exhibit a great partial current density of 74.6 mA cm⁻² and high Faradaic efficiency of 95.8%. The detailed electrocatalytic analyses and corresponding density functional theory calculations simultaneously reveal that the incorporation of Bi atoms into Sn species facilitates formate production by suppressing the formation of H₂ and CO.     

基于深度信念网络的列车故障音频诊断方法

铁路在交通运输行业有着举足轻重的地位，一旦列车发生故障将会导致严重的生命财产损失。由于列车发生故障的概率相对较低，因此难以捕获列车的故障样本。针对上述问题，提出了一种无监督学习的列车故障识别方法，通过检测列车音频信号来识别列车故障。该方法基于深度信念网络(DBN)，利用小波包分解提取检测信号的特征向量并将其作为DBN的输入，待网络充分训练后，由训练好的DBN识别当前列车的运行状况。现场监测实验结果表明，该方法能够在无监督的条件下有效识别列车故障，保障了列车的运行安全。     

基于深度学习的秀丽隐杆线虫显微图像分割方法

利用计算机实现自动、准确的秀丽隐杆线虫(C.elegans)的各项形态学参数分析,至关重要的是从显微图像上分割出线虫体态,但由于显微镜下的图像噪声较多,线虫边缘像素与周围环境相似,而且线虫的体态具有鞭毛和其他附着物需要分离,多方面因素导致设计一个鲁棒性的C.elegans分割算法仍然面临着挑战。针对这些问题,提出了一种基于深度学习的线虫分割方法,通过训练掩模区域卷积神经网络(Mask R-CNN)学习线虫形态特征实现自动分割。首先,通过改进多级特征池化将高级语义特征与低级边缘特征融合,结合大幅度软最大损失(LMSL)损失算法改进损失计算;然后,改进非极大值抑制;最后,引入全连接融合分支等方法对分割结果进行进一步优化。实验结果表明,相比原始的Mask R-CNN,该方法平均精确率(AP)提升了4.3个百分点,平均交并比(mIOU)提升了4个百分点。表明所提出的深度学习分割方法能够有效提高分割准确率,在显微图像中更加精确地分割出线虫体。     

Modeling the effect of deep traps on the capacitance–voltage characteristics of p-type Si-doped GaAs Schottky diodes grown on high index GaAs substrates

Numerical simulation, using SILVACO-TCAD, is carried out to explain experimentally observed effects of different types of deep levels on the capacitance–voltage characteristics of p-type Si-doped GaAs Schottky diodes grown on high index GaAs substrates. Two diodes were grown on (311)A and (211)A oriented GaAs substrates using Molecular Beam Epitaxy (MBE). Although, deep levels were observed in both structures, the measured capacitance–voltage characteristics show a negative differential capacitance (NDC) for the (311)A diodes, while the (211)A devices display a usual behaviour. The NDC is related to the nature and spatial distribution of the deep levels, which are characterized by the Deep Level Transient Spectroscopy (DLTS) technique. In the (311)A structure only majority deep levels (hole traps) were observed while both majority and minority deep levels were present in the (211)A diodes. The simulation, which calculates the capacitance–voltage characteristics in the absence and presence of different types of deep levels, agrees well with the experimentally observed behaviour.     

改进HSR-FCN的服装图像识别分类算法研究

目前网络上的服装图像数量增长迅猛，对于大量服装图像实现智能分类的需求日益增加。将基于区域的全卷积网络（Region-Based Fully Convolutional Networks，R-FCN）引入到服装图像识别中，针对服装图像分类中网络训练时间长、形变服装图像识别率低的问题，提出一种新颖的改进框架HSR-FCN。新框架将R-FCN中的区域建议网络和HyperNet网络相融合，改变图片特征学习方式，使得HSR-FCN可以在更短的训练时间内达到更高的准确率。在模型中引入了空间转换网络，对输入服装图像和特征图进行了空间变换及对齐，加强了对多角度服装和形变服装的特征学习。实验结果表明，改进后的HSR-FCN模型有效地加强了对形变服装图像的学习，且在训练时间更短的情况下，比原来的网络模型R-FCN平均准确率提高了大约3个百分点，达到96.69%。