考虑重叠过滤及稳定性约束的桁架拓扑优化方法

从工程实际的角度来说,一般不允许结构形式中包含重叠单元。目前桁架拓扑优化的基结构法在选定基节点的情况下一般不建重叠单元,这导致可行域缩小,使优化不能找到更优解,人为增删杆件缺乏科学依据。针对该问题,该文对重叠给出准确的数学描述,建立包含重叠杆的基结构,利用Heaviside函数将拓扑变量连续化处理,使之在优化过程中可以获取目标函数、约束函数的敏度信息,同时考虑在拓扑优化中加入基频约束以避免出现机构,并加入稳定性约束防止出现压杆失稳,通过优化模型实现重叠过滤。最后通过两个案例计算证明可以找到更优解,验证了该方法的有效性。     

实体关系抽取方法研究综述

从无结构化自然语言文本中抽取实体关系三元组是构建大型知识图谱中最为关键的一步,但现有研究仍存在3方面问题：1）忽略文本中因多个三元组共享同一实体而产生的实体关系重叠问题;2）当前以编码器−解码器为基础的联合抽取模型未充分考虑文本语句词之间的依赖关系;3）部分三元组序列过长导致误差累积与传播,影响实体关系抽取的精度和效率.基于此,提出基于图卷积增强多路解码的实体关系联合抽取模型 (graph convolution-enhanced multi-channel decoding joint entity and relation extraction model, GMCD-JERE).首先,基于BiLSTM作为模型编码器,强化文本中词的双向特征融合;其次,通过图卷积多跳特征融合句中词之间的依赖关系,提高关系抽取准确性;此外,改进传统模型按三元组先后顺序的解码机制,通过多路解码三元组机制,解决实体关系重叠问题,同时缓解三元组序列过长造成误差累积、传播的影响;最后,实验选用当前3个主流模型进行性能验证,在NYT (New York times)数据集上结果表明在精确率、召回率和F1这3个指标上分别提升了4.3%,5.1%,4.8%,同时在WebNLG (Web natural language generation)数据集上验证以关系为开始的抽取顺序.

     

A technique for deriving hyponymies and overlappings from database schemes

This paper tackles the problem of semi-automatically extracting hyponymy and overlapping properties between entities belonging to heterogeneous database schemes. The technique we propose consists of two phases: the first one derives basic hyponymies and overlappings starting from a specific situation; the second one receives basic properties derived from the first phase and extracts further, more general hyponymies. In addition, the paper reports some experimental results obtained by applying the technique presented here to the database schemes of Italian Central Government Offices. Finally, it shows some applications of derived hyponymies and overlappings; in particular it illustrates how they can influence scheme integration.     

防止异向纱线重叠的卷绕比动态调节方法

纱线卷绕中,适宜的卷绕比可以有效防止纱线间距过小所导致的重叠,但在卷绕过程中异向纱线夹角会随着纱筒直径的增大而逐渐减小,可引发纱线重叠及塌边等缺陷.为此,采用数学方法分析卷绕比与纱线夹角的微变关系,发现等差式降低卷绕比可以避免纱线夹角持续减小,同时保证了最小纱线间距适度增大且不影响纱线的排布规律.认为动态调节卷绕比可削...     

在X-射线衍射中与白云母特征峰重合的高岭土矿物组分测算

针对高岭土矿物在X-射线衍射分析过程中,组成矿物之间存在特征峰重合无法定量各组分含量的情形。该测算通过X-射线物相定量分析、化学成分分析、氧化物在矿物中所占比例等参数进行换算,计算出了高岭土各组成矿物的含量。     

A two-step cloning-free PCR-based method for the deletion of genes in the opportunistic pathogenic yeast Candida lusitaniae

We describe a new cloning-free strategy to delete genes in the opportunistic pathogenic yeast Candida lusitaniae. We first constructed two ura3 Δ strains in C. lusitaniae for their use in transformation experiments. One was deleted for the entire URA3 coding sequence; the other possessed a partial deletion within the coding region, which was used to determine the minimum amount of homology required for efficient homologous recombination by double crossing-over of a linear DNA fragment restoring URA3 expression. This amount was estimated to 200 bp on each side of the DNA fragment. These data constituted the basis of the development of a strategy to construct DNA cassettes for gene deletion by a cloning-free overlapping PCR method. Two cassettes were necessary in two successive transformation steps for the complete removal of a gene of interest. As an example, we report here the deletion of the LEU2 gene. The first cassette was constituted by the URA3 gene flanked by two large fragments (500 bp) homologous to the 5' and 3' non-coding regions of LEU2. After transformation of an ura3 Δ recipient strain and integration of the cassette at the LEU2 locus, the URA3 gene was removed by a second transformation round with a DNA cassette made by the fusion between the 5' and 3' non-coding regions of the LEU2 gene. The overall procedure takes less than 2 weeks and allows the creation of a clean null mutant that retains no foreign DNA sequence integrated in its genome.     

Study on spliced yarns of different spinning technologies. Part II: structural characteristics

A detailed study on structural parameters of splices of viscose staple fiber yarns made from ring, rotor, and air-jet spinning technologies are reported in this paper. All the yarns each had six different color tracers with 0.3% of total fibers. A three variable three-level factorial design approach proposed by Box and Behnken has been used to study the interactive effect of splicing parameters. The splicing parameters, such as splicing length, duration of splicing air blast, and splicing air pressure were varied to prepared spliced yarn samples. Splices were introduced at all these levels for the three different technologies. This paper deals with the structural characteristics of splice portions, i.e. splice diameter, helix angle, splicing coefficient of fiber (SCF), and fiber overlapping coefficient (FOC). The structural characteristics, such as splice diameter, helix angle, SCF, and FOC, are significantly affected by the different splicing parameters (i.e. splice length, duration of the opening air blast, and splicing air pressure). Different types of yarns (ring-, rotor-, and air-jet-spun) show different trends of structural characteristics with splicing parameters. Part I of this paper deals with the mechanical characteristics of spliced portion.     

自动络筒机的发展

介绍自动络筒机在定长、防叠、等张力精密卷绕、接头、上蜡、自动控制、电子清纱、节能及提高生产效率等方面的技术进步;其应用不仅为下游工序提高生产效率创造了条件,而且保障高等级纱线生产和卷绕质量,说明Autoconer 5型、POLAR型是当代最先进的全自动络筒机。     

Reduced dimension GDSW coarse spaces for monolithic Schwarz domain decomposition methods for incompressible fluid flow problems

Monolithic preconditioners for incompressible fluid flow problems can significantly improve the convergence speed compared with preconditioners based on incomplete block factorizations. However, the computational costs for the setup and the application of monolithic preconditioners are typically higher. In this article, several techniques are applied to monolithic two-level generalized Dryja-Smith-Widlund (GDSW) preconditioners to further improve the convergence speed and the computing time. In particular, reduced dimension GDSW coarse spaces, restricted and scaled versions of the first level, hybrid, and parallel coupling of the levels, and recycling strategies are investigated. Using a combination of all these improvements, for a small time-dependent Navier-Stokes problem on 240 message passing interface (MPI) ranks, a reduction of 86% of the time-to-solution can be obtained. Even without applying recycling strategies, the time-to-solution can be reduced by more than 50% for a larger steady Stokes problem on 4608 MPI ranks. For the largest problems with 11 979 MPI ranks, the scalability deteriorates drastically for the monolithic GDSW coarse space. On the other hand, using the reduced dimension coarse spaces, good scalability up to 11 979 MPI ranks, which corresponds to the largest problem configuration fitting on the employed supercomputer, could be achieved.     

优化分类模块和估计模块的快速鲁棒目标跟踪方法

为了平衡跟踪器的分类与估计模块间的性能差距,提出一种模块性能均衡的跟踪器.首先,通过大量离线学习,将高阶特征纳入目标估计中;然后,对目标估计模块进行训练,预测目标对象与估计跟踪框之间的重叠.为了提高所提分类模块在面对干扰物时的鲁棒性,引入在线训练的分类模块,采用了难分样本挖掘方法,确保较高的区分能力.在OTBI00、VOT2016和TrackingNet三个公开数据集上进行实验,结果表明,与相关滤波、多域网络等方法相比,所提方法的重叠精度指标更优、跟踪精度更高,且收敛性明显快于梯度下降法,运行速率达到实时要求.