融合字词模型的中文命名实体识别研究

命名实体识别(NER)是自然语言处理中一项非常重要的基础任务。传统的机器学习方法在处理该任务时,主要依赖于人们的专业领域知识和人工提取的特征。为了在不需要人工特征的条件下获得较好的结果,该文提出了一种融合字词BiLSTM模型的命名实体识别方法。首先分别用BiLSTM-CRF训练得到基于字的模型Char-NER和基于词的模型Word-NER,然后将两个模型得到的分值向量进行运算和拼接,将拼接后的向量作为特征送入SVM进行训练,使用SVM对Char-NER和Word-NER进行模型融合。实验结果表明,该方法在不需要人工特征的条件下,在1998年《人民日报》语料和MSRA语料上对人名、地名、机构名识别的F值分别达到了94.04%、92.15%、87.05%和91.73%、93.20%、83.15%。     

Efficient nickel catalyst with preferred orientation and microsphere for direct borohydride fuel cell

Ni catalysts with preferred orientation and microsphere were fabricated by a facile one-step pulsed voltage electrodeposition (PVE) method, while the crystal plane orientation and catalytic activity for borohydride oxidation reaction (BOR) were researched systematically. The results indicate that adjusting pulse anode potential can effectively regulate the morphology and crystal plane preferred orientation of Ni catalysts. The optimal Ni-PVE_0.15 catalyst exhibits the highest catalytic activity, stability, and fuel efficiency to BOR. The improvement of catalytic activity of Ni and fuel efficiency is attributed to the more exposure of the (220) crystal plane, which blocks the competing hydrogen evolution reaction (HER), increases the intrinsic activity of the catalytic sites to BOR, and strengthens the adsorption of escaping hydrogen to further electrooxidation. Meanwhile, the no-cracks feature of the microsphere increases the active sites and avoids the shedding of catalysts. The derived insights are important for the design of efficient nickel catalysts for practical applications of direct borohydride fuel cells.     

Deep learning technique for process fault detection and diagnosis in the presence of incomplete data

In modern industrial processes, timely detection and diagnosis of process abnormalities are critical for monitoring process operations. Various fault detection and diagnosis (FDD) methods have been proposed and implemented, the performance of which, however, could be drastically influenced by the common presence of incomplete or missing data in real industrial scenarios. This paper presents a new FDD approach based on an incomplete data imputation technique for process fault recognition. It employs the modified stacked autoencoder, a deep learning structure, in the phase of incomplete data treatment, and classifies data representations rather than the imputed complete data in the phase of fault identification. A benchmark process, the Tennessee Eastman process, is employed to illustrate the effectiveness and applicability of the proposed method.     

Natural Fragmentation Behavior of Steel Cylinders with Variable Charge Geometries under Detonation Loading

Combustion, Explosion, and Shock Waves - Natural fragmentation of steel cylinders with different charge sizes is investigated, and the correlation of different cylinders is proposed. The Mott...     

承压圆筒爆破强度预测公式的精度和稳定性

为给选择公式或拓展公式应用范围提供依据,应用概率论知识和数理统计理论,建立了公式稳定性与精度的比较与评价方法.基于承压圆筒实际爆破压力与有关公式计算值之比是基本符合正态分布随机变量的研究,从分布参数波动范围的变化系数比较与评价公式稳定性,从分布参数取值区间的重合度比较与评价公式精度.在双侧置信度98％时和公式的应用范围内,比较与评价了中径公式、特雷斯卡公式、福贝尔公式与流变应力公式的稳定性和精度.研究表明:(1)稳定性与精度不降低是选择公式或拓展公式应用范围的基本条件;(2)对于单层承压圆筒,中径公式的实测爆破压力应用范围,可由标准规定的不超过105 MPa拓展到不超过329.6 MPa,公式的稳定性明显提高且精度不降低;特雷斯卡公式的实测爆破压力应用范围,可由标准规定的位于91.0 MPa～300 MPa之间拓展到不超过329.6 MPa,公式的稳定性明显提高且精度不降低;在实测爆破压力不超过329.6 MPa时,中径公式与特雷斯卡公式的稳定性与精度无显著差异;在流变应力公式的应用范围为实测爆破压力不低于220 MPa,以及福贝尔公式的应用范围为实测爆破压力不低于250 MPa时,两个公式的精度没有明显差异,流变应力公式的稳定性明显比福贝尔公式好;当中径公式或特雷斯卡公式应用范围为实测爆破压力不超过329.6 MPa,以及流变应力公式的应用范围为实测爆破压力不低于220 MPa时,中径公式或特雷斯卡公式的稳定性明显比流变应力公式的好,精度也高于流变应力公式.(3)对于实测爆破压力不超过209.7 MPa的扁平绕带式压力容器,中径公式的精度和稳定性比单层承压圆筒爆破压力的4个预测公式低.     

Three-Dimensional Normal Stress for Controlling Electronic Structure and Magnetic Property of Fe2 Ge

A system study of the three-dimensional normal stress for regulating electronic structure and magnetic property of Fe_2Ge is studied. The density states of Fe more than 92% contribution come from Fe 3d,the density states of Ge mainly contributed from Ge 4p and Ge 4s,and the Fe 3d spin induces the Ge 4p electron transfer. The inductive effect increases germanium electron energy,weakens the Fe spin density of states,opposes the stability of the ferromagnetic state. The magnetic moment varies from 5 to 3 μB with the stress charges from-30 to 30 GPa. The charge of Fe is negative whereas the Ge atom is positively charged,the Fe atom loses charge,the charge transfers to the Ge atom. The unevenly distributed charge forms the newoccupy state and spin polarization state in the Fe_2Ge electron structure system. The Fe is the electron donor,the total electron is transferred to Ge,but the total numbers of gain electron and total numbers of lost electron are not equal,so the Fe_2Ge electron system may have hybridization between the Fe 3d state and Ge 4p state.The magnetic of Fe_2Ge mainly comes from the unoccupied Fe 3d orbital,the Fe 3d is positive spinpolarization state and the spin-polarization strength is decreased,the Ge 4p is negative spin-polarization state and the spin-polarization strength are increased. M oreover,electrons-spin polarization is relevant to the structure parameters of the Fe_2Ge system,and controls spin-polarized electronic behavior by means of adjusting ferromagnetic.     

Effect of Cooling Rate on the Microstructure of Semi-Solid Al–25Si–2Fe Alloy During Electromagnetic Stirring

The effect of cooling rate on the microstructure of semi-solid Al–25Si–2Fe alloy was investigated during electromagnetic stirring. It was found that as the cooling rate was increased from 7 to 21 °C/min, the equivalent diameter of the primary Si particles decreased from 70 ± 5 to 25 ± 2 μm. The primary Si particles form a fine blocky structure when the cooling rate is 21 °C/min. When the cooling rate is increased to 30 and 33 °C/min, the primary Si particles coarsen and adopt plate or other irregular shapes. As the melt cools to 690 °C, Fe inter-metallic phases present different morphologies at different cooling rates during EMS. These phases in the Al–25Si–2Fe alloy are mainly in the form of δ-Al₄FeSi₂ at higher cooling rates.     

硫化罐卡箍的质量优化设计

应用CAE软件Proe,Ansys和协同优化软件Isight,构建了基于Isight的协同优化设计平台,解决了软件之间的数据通信技术问题和设计分析过程的自动化技术问题。并利用这一平台、基于分析设计的思想成功完成了对硫化罐卡箍的质量优化设计,为今后过程装备的改进优化提供了一种技术路线。