基于模糊决策的主旨句生成算法

提炼文档的主旨句在自动文本摘要生成系统中处于十分重要的地位.传统方法给出一个计算句子权重的公式,但此公式形式单一固定,不够灵活,通过借鉴模糊决策理论,提出了一种基于模糊决策的主旨句生成方法.该方法不局限于影响句子权重因素的数量及形式,它依据模糊决策理论对文档中每个句子的重要程度做出综合评判,将那些较大程度上反映文章主旨的语句提炼出来,作为摘要的生成语句.由于人们在主观上对各个因素的重视程度也有所不同,因此,又提出一种改进算法.实验结果表明,给出方法所提取的文章主旨语句较传统公式法的准确性大大提高了.     

关于提高变频器调频速度的研究

本文介绍了作者的一项专利技术。对于需要比较快速、比较频繁地调整变频器频率应用场合,目前的通用变频器就很难满足这个需要,尤其是在大惯性负载上使用变频器时,须将升、降频（特别是降频）速度设置的比较慢,以适应由于惯性造成的电机转速的滞后。本专利技术克服上述传统变频器目前存在的缺陷,在传统变频器上增加比较简单并业已成熟的换相技术,为变频器的制造提供了一种新的附加技术。     

Stress intensity factors for surface cracks in single‐edge notch bend specimen by a three‐dimensional weight function method

Stress intensity factors for half‐elliptical surface cracks at a semi‐circular notch in a recently developed single‐edge notch bend specimen are determined for a wide range of geometrical parameters using a three‐dimensional weight function method. Two load cases of pin loading and uniform remote tension are considered. The results are in good agreement with abaqus/franc3d finite element analysis. It is found that the Ziegler–Newman engineering similitude approach (programmed into the Fatigue Crack Growth Structural Analysis life‐prediction code) produces good results for a wide range in a/c ratios. Expressions by multi‐variable curve fitting to the weight function results are presented for easy engineering applications.     

新型模块化可重构机器人设计与运动学分析

采用模块化设计思想,设计了一种新型可重构工业机器人.通过对机器人进行结构优化设计,机器人的负载自重比达到1/4.5.机器人各关节模块的结构均相同,大幅减少了设计工作量;关节模块中传动部件均采用通用件,降低了机器人的成本.通过关节模块和手臂模块的组合,可形成多种机器人构型.针对其中一种机器人构型,利用DH法建立了机器人正运动学模型.该构型机器人不符合相邻的3根关节轴线交于一点的条件,利用传统方法很难求得全部逆解,因此提出一种代数法与几何法相结合的新方法,求得该构型机器人逆运动学的完整解析解.机器人运动学分析的正确性通过算例进行了验证,为机器人的运动控制奠定基础.     

Backcalculation of viscoelastic and nonlinear flexible pavement layer properties from falling weight deflections

Appropriate characterisation of individual layer properties is crucial for mechanistic analysis of flexible pavements. Typically in inverse analyses, pavements are modelled as elastic or nonlinear elastic to obtain layer material properties through non-destructive falling weight deflectometer (FWD) testing. In this study, a layered viscoelastic–nonlinear forward model (called LAVAN) was used to develop a genetic algorithm-based backcalculation scheme (called BACKLAVAN). The LAVAN can consider both the viscoelastic behaviour of asphalt concrete (AC) layer and nonlinear elastic behaviour of unbound layers. The BACKLAVAN algorithm uses FWD load-response history at different test temperatures to backcalculate both the (damaged) E(t) and |E^*| master curve of AC layers and the linear and nonlinear elastic moduli of unbound layers of in-service pavements. The BACKLAVAN algorithm was validated using two FWD tests run on a long-term pavement performance section. Comparison between the backcalculated and measured results indicates that it should be possible to infer linear viscoelastic properties of AC layer as well as nonlinear elastic properties of unbound layers from FWD tests.     

Combustion Waves on the Point Set Surface

This paper introduces a combustion model of heat transfer and fuel consumption for the propagation of a fire front on a point cloud surface. The heat transfer includes the heat advection by the airflow as well as diffusion, chemical reaction, and heat loss to generate complex, but controllable heat flows with a designed airflow velocity. For the stable heat advection, we solve a semi‐Lagrangian method on point samples using discrete exponential maps to trace the position from which the wind blows while preserving the geodesic distance. We also propose angular Voronoi weights for a discrete Laplace‐Beltrami operator that shows better isotropic diffusion on the inhomogeneous distribution of point clouds than the cotangent or moving least‐squares schemes. We demonstrate a diversity of burning scenarios by incorporating factors affecting the fire spreading such as buoyancy and object geometries in the airflow velocity fields, or by synthesizing patterns.     

碾压混凝土现场生产性试验

介绍某水电站大坝浇筑碾压混凝土现场生产性试验,通过对不同VC值和碾压遍数试验确定VC值和碾压遍数;采用不同层面处理方法试验,确定层面处理工艺;根据不同的注浆方式试验,确定变态混凝土注浆工艺。     

基于生存能力的节点权值多样化布置算法

将多样化应用到网络系统、软件中以减少恶意代码传播以及单一重复攻击的多米诺效应是一个广泛的共识.提出一种基于生存能力的节点权值多样化布置算法,从节点重要性出发,将有限的差异系统布置到所有节点中.该算法能在保证单个节点对网络影响最小前提下,实现核心节点的高生存性.仿真实验表明了算法的有效性.     

基于模糊粗糙集和D-S证据理论的多源灌溉信息融合方法

针对多源灌溉信息决策过程中不确定性信息难以融合的问题,提出了一种基于模糊粗糙集和D-S证据理论相结合的决策融合方法。运用模糊粗糙集理论,建立基本概率分配函数,计算各灌溉因子与灌溉决策之间的依赖程度,构建多个融合灌溉因子对灌溉决策的识别框架;然后运用改进的D-S证据理论,进行多源灌溉信息决策层级的融合,最终解决不确定信息的表达和合成问题。应用上述方法对华北地区冬小麦土壤水分、光合速率和气孔导度等信息进行灌溉决策融合,结果显示：灌溉决策的不确定性由融合前的最高38%降至9.84%,该方法可有效地提高灌溉决策精度,降低灌溉决策的不确定性