基于改进VGG网络的单体热电池X光图像无损检测方法研究

为解决单体热电池生产中出现的安装错误、人工检测耗时耗力的问题,提出一个结合迁移学习和卷积 神 经网络(convolutional neural network,CNN) 的单体热电池缺陷检测模型。首先,对数据集图像进行裁剪、加噪等预处理,以VGG16(visual geometry group 16) 网络作为 模型的骨干架构,在瓶颈层后增添选择性核(selective kernel,SK) 卷积;然后,增添全局平均池化(global average pooling,GAP) 层, 增加Dropout层及添加 L2 正则化等微调操作,得到单体热电池缺陷检测模型Q-VGGNet;最后,在大型公开数据集ImageNet上进 行预训练学习,将获得的权重参数迁移到单体热电池图像识别模型Q-VGGNet上。测试实验表明:6种 网络模型对数据集缺陷图像的总体识别准确率分别达到了98.39%、94.44%、97.27%、96.34%、93.71%、 95.61%,Q-VGGNet网 络模型 对合格图像和 漏装负极、极耳断裂、漏装集流片3种缺陷图像 识别准确率 分别达到了99.6%,95.9%,99.6%和98.4%。检测结果表明:该方法能够更准确、快速地检测热电池缺陷, 拥有良好的缺陷诊断能力,较传统方法提高近3%,为人工检测单体热电池缺陷提供了良好的解决途径。     

融合注意力线性特征多样化的DR分级模型

糖尿病视网膜病变(diabetic retinopathy, DR)是目前人类的主要致盲疾病之一。针对DR数据集中样本类间差异小和类分布不均衡等制约分级性能提高的问题,本文提出一种融合注意力线性特征多样化(fusion of attention linear feature diversification, FALFD)的分级算法。该算法首先用改进的Res2Net残差网络作为模型骨干来增大感受野,进一步提高网络捕捉特征信息的能力;其次引入自适应特征多样化模块(adaptive feature diversification module, AFDM)对眼底图像可分辨的微小病理特征进行识别,获得具有高语义信息的局部特征,避免单一特征区域的限制,进而提高分级准确度;再后利用双线性注意力融合模块(bilinear attention fusion module, BAFM)增加可判别区域特征的网络权重占比;最后采用正则化焦点损失(focal loss, FL)进一步提升算法的分类性能。在IDRID数据集上,灵敏度和特异性分别为94.20%和97.05%,二次加权系数为87.83%;在APTO...     

基于时域有限元的近场单站导体柱微波成像

介绍了一种在自由空间中对导体柱近场单站电磁成像的时域数值方法.提出以不均匀分布的点--离散点描述法,来代替通常采用的有限项三角级数逼近的形状函数来描述散射体.在正过程中采用时域有限元法获得其近区散射场,逆过程则通过遗传算法来寻找最优解.以不均匀分布于导体表面点的矢径的模作为优化变量.以在整个迭代过程中,目标散射场和计算散射场的误差总和与目标散射场场值绝对值之和的比值作为目标函数,使目标函数达到最小值来获得导体柱在自由空间中的电磁成像.     

Anisotropic Superconducting Properties of Kagome Metal CsV_3Sb_5

We systematically measure the superconducting(SC) and mixed state properties of high-quality CsV_3 Sb_5 single crystals with T_c～3.5 K.We find that the upper critical field H_(c2)(T) exhibits a large anisotropic ratio of H_(c2)~(ab)/H_(c2)~c～9 at zero temperature and fitting its temperature dependence requires a minimum two-band effective model.Moreover,the ratio of the lower critical field,H_(c1)~(ab)/H_(c1)~c,is also found to be larger than 1,which indicates that the in-plane energy dispersion is strongly renormalized near Fermi energy.Both H_(c1)(T) and SC diamagnetic signal are found to change little initially below T_c～3.5 K and then to increase abruptly upon cooling to a characteristic temperature of ～2.8 K.Furthermore,we identify a two-fold anisotropy of in-plane angular-dependent magnetoresistance in the mixed state.Interestingly,we find that,below the same characteristic T～2.8 K,the orientation of this two-fold anisotropy displays a peculiar twist by an angle of 60° characteristic of the Kagome geometry.Our results suggest an intriguing superconducting state emerging in the complex environment of Kagome lattice,which,at least,is partially driven by electron-electron correlation.     

诺基亚西门子通信GPRS网络容量分析体系的研究

本文通过对中国移动通信集团新疆有限公司的GPRS网络结构及容量需求分析入手,提出了容量备份方案,有效解决了业务发展带来的容量不足问题,并且通过网络结构的调整,使得GPRS网络更加安全可靠,大幅度提升了网络的健壮性,为业务高效稳定的发展奠定了基础。     

Unsourced Multiple Access for 6G Massive Machine Type Communications

Unsourced multiple access(UMA) is a multi-access technology for massive, low-power, uncoordinated, and unsourced Machine Type Communication(MTC) networks. It ensures transmission reliability under the premise of high energy efficiency.Based on the analysis of the 6G MTC key performance indicators(KPIs) and scenario characteristics,this paper summarizes its requirements for radio access networks. Following this, the existing multiple access models are analyzed under these standards to determine U...     

Implementation Framework and Validation of Cluster-Nuclei Based Channel Model Using Environmental Mapping for 6G Communication Systems

With the research of the upcoming sixth generation(6 G) systems, new technologies will require wider bandwidth, larger scale antenna arrays and more diverse wireless communication scenarios on the future channel modeling. Considering channel model is prerequisite for system design and performance evaluation of 6 G technologies, we face a challenging task: how to accurately and efficiently model 6 G channel for various scenarios? This paper tries to answer it. Firstly, the features of cluster-nuc...     

Antibacterial Activity of Aureonuclemycin Produced by Streptomyces aureus Strain SPRI-371

Actinomycetes play a vital role as one of the most important natural resources for both pharmaceutical and agricultural applications. The actinomycete strain SPRI-371, isolated from soil collected in Jiangsu province, China, was classified as Streptomyces aureus based on its morphological, physiological, biochemical and molecular biological characteristics. Its bacterial activity metabolites were identified as aureonuclemycin (ANM), belonging to adenosine derivatives with the molecular formula C₁₆H₁₉N₅O₉ for ANM A and C₁₀H₁₃N₅O₃ for ANM B. Simultaneously, the industrial fermentation process of a mutated S. aureus strain SPRI-371 was optimized in a 20 m³ fermentation tank, featuring a rotation speed of 170 rpm, a pressure of 0.05 MPa, an inoculum age of 36–40 h and a dissolved oxygen level maintained at 1–30% within 40–80 h and at >60% in the later period, resulting in an ANM yield of >3700 mg/L. In the industrial separation of fermentation broth, the sulfuric acid solution was selected to adjust pH and 4# resin was used for adsorption. Then, it was resolved with 20% ethanol solution and concentrated in a vacuum (60–65 °C), with excellent results. Antibacterial experiments showed that ANM was less active or inactive against Xanthomonas oryzae pv. oryzae, Xanthomonas citri subsp. citri and Xanthomonas oryzae pv. oryzicola and most bacteria, yeast and fungi in vitro. However, in vivo experiments showed that ANM exhibited extremely significant protective and therapeutic activity against diseases caused by X. oryzae pv. oryzae and X. oryzae pv. oryzicola in rice and X. citri in oranges and lemons. In field trials, ANM A 150 gai/ha + ANM B 75 gai/ha exhibited excellent therapeutic activity against rice bacterial leaf blight, citrus canker and rice bacterial leaf streak. Furthermore, as the dosage and production cost of ANM are lower than those of commercial drugs, it has good application prospects.     

Adaptive coordination assemblies based on a flexible tetraazacyclododecane ligand for promoting carbon dioxide fixation

Coordination hosts based on flexible ligands have received increasing attention due to their inherent adaptive cavities that often show induced-fit guest binding and catalysis like enzymes. Herein, we report the controlled self-assembly of a series of homo/heterometallic coordination hosts (Me₄enPd)_2n(ML)_n [n = 2/3; M = Zn(ii)/Co(ii)/Ni(ii)/Cu(ii)/Pd(ii)/Ag(i); Me₄en: N,N,N′,N′-tetramethylethylenediamine] with different shapes (tube/cage) from a flexible tetraazacyclododecane-based pyridinyl ligand (L) and cis-blocking Me₄enPd(ii) units. While the Ag(i)-metalated ligand (AgL) gave rise to the formation of a (Me₄enPd)₄(ML)₂-type cage, all other M(ii) ions led to isostructural (Me₄enPd)₆(ML)₃-type tubular complexes. Structural transformations between cages and tubes could be realized through transmetalation of the ligand. The buffering effect on the ML panels endows the coordination tubes with remarkable acid–base resistance, which makes the (Me₄enPd)₆(ZnL)₃ host an effective catalyst for the CO₂ to CO₃²⁻ conversion. Control experiments suggested that the integration of multiple active Zn(ii) sites on the tubular host and the perfect geometry match between CO₃²⁻ and the cavity synergistically promoted such a conversion. Our results provide an important strategy for the design of adaptive coordination hosts to achieve efficient carbon fixation.

A series of coordination hosts were prepared and their applications in CO₂ fixation were studied.