氧化铟-稀土氧化物共掺CeO2基导电电解质材料的制备与表征

设计了以氧化铟与稀土氧化物（La2O3或Nd2O3）为共同掺杂元素的新型掺杂技术，并利用以聚乙烯醇为聚合剂的化学合成方法制备了CeO2基氧离子导电电解质材料。X射线衍射分析表明，本实验采用的湿化学合成方法容易获得具有纳米结构的高纯氟化钙型结构的CeO2基导电材料．材料导电性的交流阻抗测试分析说明，适当比例的氧化铟与稀土氧化物共掺能有效提高CeO2的导电性；其导电性比相同合成方法制备的Sm2O3掺杂CeO2的导电性更好。利用有效离子半径和相关结合焓理论分析了不同掺杂氧化物对CeO2基材料离子导电性影响。     

Conductivity of cerium oxides doped by trivalent cations

For the doping technique with single dopant of Sb2O3, Ga2O3, In2O3, AlCl3, Bi2O3 or Sm2O3, the dopant concentrations of 10%（molar fraction） were selected for the CeO2-based oxygen-ion conductors. A wet chemical method by polymerization of polyvinyl-alcohol was performed for the compound preparation. X-ray diffraction analysis result shows that the calcined powders have nanocrystalline structure with grain size less than 40 nm, and the sintered samples have pure single fluorite-type phase in all doped cerium oxides. Impedance measurement results show that the conductivities and activation energies of CeO2-based oxides depend closely on the dopants. The Sb2O3-doped CeO2 has the largest conductivity and In2O3-doped CeO2 has the lowest one at 200℃, the Bi2O3-doped oxide shows the highest activation energy and Al203-doped oxide shows the lowest one. Ionic radii and association enthalpies of dopants were taken account for different conductivities and activation energies of the CeO2-based oxygen-ion conductors.     

基于光通路状态感知的分簇式故障定位机制

针对现有故障定位机制定位时间长和对业务分布依赖高等问题,该文提出基于光通路状态感知的分簇式故障定位机制。该机制根据网络分簇约束条件,以最小支配集理论为基础,建立两级网络模型。并且根据算法特点,定义了适用于该算法的矩阵与运算。故障后簇头节点以及汇聚节点通过对各节点发送的矩阵进行矩阵与运算实现快速准确的故障定位。仿真表明,该机制以较低的复杂度和资源开销,有效地降低了对业务分布的依赖,极大地提升了故障定位率,减少了故障定位时间。     

Crystal preferential orientation ofAl-1.0%Si alloy in continuous unidirectional solidification process

Al-1.0%Si alloy rods containing unidirectional microstructures were fabricated by continuous unidirectional solidification （COS） process. The crystal growth orientation evolution of Al-1.0%Si alloy in CUS process was studied. The microstructure was observed and the orientation results examined by electron back scattering diffraction（EBSD） indicate that at the beginning of COS process the crystals grow along a certain preferred orientation, （100）direction, but there are also other growth directions deflecting from this preferred orientation. It is found that as directional solidification continues, the crystals with high growth speed remain and those with low growth speed vanish. The crystal preferred orientation tends to be （100）direction in competition growth process. Namely, the （100） fiber texture oral-1.0%Si alloy rod becomes stronger in the stable directional solidification establishing process. When the solidification parameters remain stable, the crystal growth orientation also obtains in steady state. Once the technological parameters fluctuate, the stray crystals deviating from the preferred orientation appear. After the process of competition growth the crystals regain strong preferred orientation. The orientation distribution function（ODF） results also testify the above conclusions.     

基于多粒度特征表示的知识图谱问答

近年来,基于知识图谱的问答系统逐渐成为学术界和工业界的研究和应用热点方向,而传统方法通常存在效率不高以及未充分利用数据信息的问题。针对以上问题,本文将中文知识图谱问答分为实体抽取和属性选择2个子任务,采用双向长短期记忆条件随机场（Bi-LSTM-CRF）模型来进行实体识别,并提出一种多粒度特征表示的属性选择模型。该模型采用字符级别以及词级别分别对问句和属性进行嵌入表示并通过编码器进行编码,对于属性同时还引入热度编码的信息。通过不同粒度文本表示的结合,并对问句和属性进行相似度计算,最终该系统在NLPCC-ICCPOL 2016 KBQA数据集上取得了73.96%的F1值,能够较好地完成知识图谱问答任务。