增强神经网络辨识模型泛化能力的研究

神经网络(Artificial Neural Network,ANN)辨识模型的泛化能力是其最主要的性能之一,增强ANN模型的泛化能力也是近年来国内外有关专家学者研究的重点问题。大量研究表明,ANN模型泛化能力的改善与很多因素相关联,其中恰当的性能指标函数设计是一个重要影响因素。文中在分析常见的基于均方误差最小原则的性能指标函数基础上,通过加入ANN辨识模型权值间的延迟信息,进而获得一种改进型性能指标函数。通过仿真,验证了所设计的改进型性能指标函数对增强ANN辨识模型的泛化能力是有效的。     

干式铣削Al-50wt%Si高硅铝合金刀具磨损与表面粗糙度评价

本文以Al-50wt%Si高硅铝合金为研究对象，采用单因素试验方法进行无涂层硬质合金刀具干式铣削试验，分析切削参数对刀具磨损和表面粗糙度的影响。结果表明：表面粗糙度受每齿进给量的影响最显著，随每齿进给量的增加而增加，当每齿进给量从0.07 mm/z增加到0.16 mm/z时，表面粗糙度增加2倍；刀具磨损受切削速度的影响最显著，随切削速度的增加而增加，当切削速度从140 m/min增加到260 m/min时，切削总长度降低3倍，而刀具后刀面磨损量仅是260 m/min速度下的0.8倍；表面粗糙度随刀具磨损的增加呈现先增加后降低的变化趋势，切削长度从350 mm增加到1 750 mm，刀具磨损量平均增加4.5倍，而表面粗糙度却下降2倍；硬质合金刀具主要的磨损形式为磨粒磨损、崩刃。     

Parameter optimization for torsion spring of deployable solar array system with multiple clearance joints considering rigid–flexible coupling dynamics

In this paper, four novel evaluation indices and corresponding hierarchical optimization strategies are proposed for a deployable solar array system considering panel flexibility and joint clearance. The deployable solar array model consists of a rigid main-body, two panels and four key mechanisms, containing torsion spring mechanism, closed cable loop mechanism, latch mechanism and attitude adjustment mechanism. Rigid and flexible components are established by Nodal Coordinate Formulation and A...     

Optimum alternate material selection methodology for an aircraft skin

Weight penalty has been a challenge for design engineers of aerospace vehicles. Today’s high-efficiency combat aircraft undergoes intense stress and strain during flying missions, which require stronger and stiffer materials to retain structural integrity. Though metallic materials have been successfully used for the construction of aircraft structures and components, metals still have a low strength-to-weight ratio. This paper aims to develop an alternate optimised material selection methodolog...     

Effects of surface topography and specimen thickness on high-speed raindrop impact damage of CFRP laminates

Raindrop impact erosion has been observed since early days of aviation, and can be catastrophic for exposed materials during supersonic flight. A single impact waterjet apparatus is established for mimicking drop impacts at the velocities between 350 m/s and 620 m/s. Carbon Fiber Reinforced Polymer(CFRP) laminates with three different surface morphologies and specimen thicknesses are tested here. A central region with no visible damage has been noticed, surrounded by a “failure ring” with common...     

一种针对多光谱遥感图像的桥梁快速识别算法

提出了一种针对多光谱图像中桥梁的识别算法。首先,根据水体和背景地物在不同光谱波段的亮度差异,计算多光谱图像的水体指数得到水体增强图,搜索其具有明显双峰的直方图得到最优阈值,实现河流的完整提取；其次,利用桥梁的存在会导致局部水体的光谱异常,沿河流中间线进行潜在桥梁区域的快速提取；再进一步利用桥梁长度以及与河流的空间关系进行鉴别,有效剔除虚警。利用 SPOT4遥感影像进行实验,结果表明本文算法运算量小,对于多个桥梁的识别具有很好的实用性。     

固态激光武器系统指标设计与分析

针对固态激光武器技术的发展,利用其非相干合束的能量叠加作用机理,通过将激光束衍射、光束质量、湍流、跟瞄误差等因素归一化成远场光斑的扩展,对子口径发射时的指标进行仿真分析,并根据仿真结果、子口径与系统发射口径间的关系,对固态激光武器系统的单台激光器功率、合束数量、发射口径,以及光束质量、跟瞄误差等因素进行了综合分析,给出了一定输入条件下的武器系统指标备选方案。     

宜香1A及宜香品种成熟期倒伏指数研究

以籼型三系不育系宜香1A、Ⅱ-32A、冈46A和恢复系宜恢7633、宜恢7008、宜恢1525等为材料,研究了亲本和杂交后代的倒伏指数.结果表明,Ⅱ-32A的倒3节间抗倒能力极显著强于宜香1A;冈46A倒3节间的抗倒能力显著强于宜香1A;宜香1A和Ⅱ-32A所配品种倒3节间抗倒能力相当,都极显著优于冈46A 所配品种倒3节间抗倒能力.母本及其所配品种的倒2节间和倒1节间倒伏指数无明显差异.