深度优先局部聚合哈希

已有的深度监督哈希方法不能有效地利用提取到的卷积特征,同时,也忽视了数据对之间相似性信息分布对于哈希网络的作用,最终导致学到的哈希编码之间的区分性不足.为了解决该问题,提出了一种新颖的深度监督哈希方法,称之为深度优先局部聚合哈希(Deep Priority Local Aggregated Hashing,DPLAH).DPLAH将局部聚合描述子向量嵌入到哈希网络中,提高网络对同类数据的表达能力,并且通过在数据对之间施加不同权重,从而减少相似性信息分布倾斜对哈希网络的影响.利用Pytorch深度框架进行DPLAH实验,使用NetVLAD层对Resnet18网络模型输出的卷积特征进行聚合,将聚合得到的特征进行哈希编码学习.在CI-FAR-10和NUS-WIDE数据集上的图像检索实验表明,与使用手工特征和卷积神经网络特征的非深度哈希学习算法的最好结果相比,DPLAH的平均准确率均值要高出11％,同时,DPLAH的平均准确率均值比非对称深度监督哈希方法高出2％.     

Coupling effect of the conductivities of Li ions and electrons by introducing LLTO@C fibers in the LiNi0.8Co0.15Al0.05O2 cathode

To probe the coupling effect of the electron and Li ion conductivities in Ni-rich layered materials(LiNi0.8Co0.15Al0.05O2,NCA),lithium lanthanum titanate(LLTO)nanofiber and carbon-coated LLTO fiber(LLTO@C)materials were introduced to polyvinylidene difluoride in a cathode.The enhancement of the conductivity was indicated by the suppressed impedance and polarization.At 1 and 5 C,the cathodes with coupling conductive paths had a more stable cycling performance.The coupling mechanism was analyzed based on the chemical state and structure evolution of NCA after cycling for 200 cycles at 5 C.In the pristine cathode,the propagation of lattice damaged regions,which consist of high-density edge-dislocation walls,destroyed the bulk integrity of NCA.In addition,the formation of a rock-salt phase on the surface of NCA caused a capacity loss.In contrast,in the LLTO@C modified cathode,although the formation of dislocation-driven atomic lattice broken regions and cation mixing occurred,they were limited to a scale of several atoms,which retarded the generation of the rock-salt phase and resulted in a pre-eminent capacity retention.Only NiO phase“pitting”occurred.A mechanism based on the synergistic transport of Li ions and electrons was proposed.     

Dynamic mass variation and multiphase interaction among steel,slag, lining refractory and nonmetallic inclusions: Laboratory experiments and mathematical prediction

The mass transfer among the multiphase interactions among the steel, slag, lining refractory, and nonmetallic inclusions during the refining process of a bearing steel was studied using laboratory experiments and numerical kinetic prediction. Experiments on the system with and without the slag phase were carried out to evaluate the influence of the refractory and the slag on the mass transfer. A mathematical model coupled the ion and molecule coexistence theory, coupled-reaction model, and the surface renewal theory was established to predict the dy-namic mass transfer and composition transformation of the steel, the slag, and nonmetallic inclusions in the steel. During the refining process, Al2O3 inclusions transformed into MgO inclusions owing to the mass transfer of [Mg] at the steel/refractory interface and (MgO) at the slag/re-fractory interface. Most of the aluminum involved in the transport entered the slag and a small part of the aluminum transferred to lining re-fractory, forming the Al2O3 or MgO·Al2O3. The slag had a significant acceleration effect on the mass transfer. The mass transfer rate (or the re-action rate) of the system with the slag was approximately 5 times larger than that of the system without the slag. In the first 20 min of the re-fining, rates of magnesium mass transfer at the steel/inclusion interface, steel/refractory interface, and steel/slag interface were x, 1.1x, and 2.2x, respectively. The composition transformation of inclusions and the mass transfer of magnesium and aluminum in the steel were predicted with an acceptable accuracy using the established kinetic model.     

用于高性能超级电容器的基于蚕沙的二氧化锰/碳纳米复合材料

MnO₂/biomass carbon nanocomposite was synthesized by a facile hydrothermal reaction. Silkworm excrement acted as a carbon precursor, which was activated by ZnCl₂ and FeCl₃ combining chemical agents under Ar atmosphere. Thin and flower-like MnO₂ nanowires were in-situ anchored on the surface of the biomass carbon. The biomass carbon not only offered high conductivity and good structural stability but also relieved the large volume expansion during the charge/discharge process. The obtained MnO₂/biomass carbon nanocomposite electrode exhibited a high specific capacitance (238 F·g?1 at 0.5 A·g?1) and a superior cycling stability with only 7% degradation after 2000 cycles. The observed good electrochemical performance is accredited to the materials’ high specific surface area, multilevel hierarchical structure, and good conductivity. This study proposes a promising method that utilizes biological waste and broadens MnO₂-based electrode material application for next-generation energy storage and conversion devices.     

低轨星座导航增强能力研究与仿真

低轨卫星星座具有几何图形变化快、落地信号功率强、全球天基监测覆盖等天然优势,可对中高轨全球导航卫星系统(Global Navigation Satellite System,GNSS)星座进行有效补充和增强,提升全球定位、导航与授时(Positioning,Navigation and Timing,PNT)服务的精度、完好性、可用性和抗干扰等能力,已成为下一代卫星导航系统重要的发展方向.本文总结了国内外低轨星座发展现状,对不同低轨星座进行了分析和设计,对低轨星座提升导航定位精度、加速精密单点定位(Precise Point Positioning,PPP)收敛、全球天基监测等导航增强能力进行了分析,重点论证了低轨星座突破现有中高轨GNSS技术瓶颈的机遇和体系增量能力,给出了相应的仿真分析结果,以期对我国低轨星座和北斗卫星导航系统的融合发展提供参考和建议.     

江淮东部平原晚更新世末植被演替及气候变化

在对位于江淮东部平原的江苏兴化蒋家舍地层剖面进行野外考察采样、AMS14 C年代测定和孢粉分析的基础上,初步探讨了研究区晚更新世末的植被演替及气候变化.结果表明:1)27.9-13.5cal.ka BP,剖面中孢粉含量高,植物的科属也丰富.其中,27.9-26.7cal.ka BP、14.5-14.2 cal.ka BP时段研究区植被为针阔叶混交林,反映气候温暖较湿润,而26.7-14.5 cal.ka BP期间地层出现沉积间断,气候寒冷干燥;14.2-13.5 cal.ka BP时段草本植物含量较高,总体为针阔叶混交林草原植被,气候温凉稍干.2)13.5-12.6cal.ka BP,剖面中孢粉含量较低,植物的科属也较贫乏,木本植物所占百分比较低,草本植物花粉含量相对较高.其中,13.5-12.8 cal.ka BP为覆盖度趋低的针阔叶混交林或草原植被,气候向干冷发展,12.8-12.6 cal.ka BP时段为针阔叶混交林植被,木本植物含量有所上升,反映气候稍暖湿.3)针叶乔木松属花粉和蕨类植物水蕨科植物孢子在剖面中自下而上均有不同程度的分布,常绿阔叶和落叶阔叶类出现率较低,这种现象跟当地现代植被的植物组合略有差异.     

一种实用的大功率光伏逆变器窄脉冲抑制方法

针对窄脉冲会严重影响大功率逆变器系统可靠运行的普遍现象,在分析窄脉冲的危害及其产生原理基础上,指出了影响窄脉冲产生的主要因素是调制比大小,据此提出一种在脉冲宽度调制(PWM)递增递减模式下的窄脉冲抑制策略,并进行实验验证。结果表明,采用的窄脉冲抑制算法能有效抑制窄脉冲信号,避免窄脉冲对系统控制的危害,提高系统控制的稳定性和安全性。     

地铁车站小直径管幕工法开挖变形规律

沈阳地铁市府大路站是采用小直径管幕工法施工的地铁暗挖车站.通过有限差分软件FLAC^3D建立车站结构-地层三维模型,分析总结了小直径管幕工法动态开挖过程引起的地层及主体结构的位移变形规律.结果表明:小直径管幕工法开挖过程引起的地表沉降具有明显的阶段性,沉降槽形态在群洞效应和管幕预支护作用的影响下变化频繁,横导洞间土体开挖阶段引起的地表沉降占到了最终沉降的50.54%,该阶段是地表沉降控制的关键阶段;管幕-梁-桩-柱施工完成后,主体结构变形表现出良好的协同作用,横梁竖向变形表现为车站两端小中间大,边桩及中柱的水平位移在负二层施工期间增长显著,占到了水平位移最大值的47.1%和55.8%,该阶段是控制主体结构变形的关键阶段.     

基于海林格距离加权关键主元的流程工业故障检测研究

在采用主成分分析（principal component analysis,PCA）算法进行故障检测时,主元的选取及处理直接影响其故障检测的表现。对此,提出一种基于全变量表达（full variable expression,FVE）和海林格距离（Hellinger distance,HD）的故障检测方法。首先,利用FVE得到所有关键主元,即保留所有变量信息;然后考虑到与故障相关主元的重要性,定义基于海林格距离的变化率,用来衡量正常工况下主元与异常工况下主元的差异;对与故障发生更相关的主元进行加权,以突出与故障相关主元对于后续故障检测的影响;最后,考虑到降维后数据通常服从非高斯分布,利用改进的局部离群因子（local outlier factor,LOF）构建统计量,其相应控制限通过核密度估计（kernel density estimation,KDE）确定。数值实例及带钢热连轧实际生产数据验证了所提方法的有效性与优越性。     

酵母耐高温分子机制的研究进展

耐高温是酵母在工业生产中的重要特性之一.近年来,人们对耐高温酵母的应用研究日益重视并取得较大的进展,但其耐高温分子机制尚未明确.目前的研究结果表明,酵母耐高温机制涉及代谢网络调控、信号转导途径及多基因/蛋白共同作用等方面.本文从代谢途径、基因表达、酶活特性和蛋白质相互作用等4个方面综述当前酵母耐高温分子机制的研究进展,...