基于支持向量机的组合分类方法及应用

为了解决采用神经网络、决策树作为弱分类器的AdaBoost组合分类存在的不足,进一步改善组合分类效果,提出采用支持向量机(SVM)作为弱分类器的一种新的组合分类诊断方法——AdaBoost-SVM。该方法没有采用一个固定的SVM的核参数,而是自适应调整SVM中的核参数,从而得到一组有效的SVM弱分类器。通过对基准数据库的测试及航空发动机故障样本的诊断,结果表明,所提AdaBoost-SVM方法较好地解决了现有的Ada-Boost组合分类方法中存在的弱分类器本身参数选取困难问题及训练轮数的合理选取问题,并具有更好的泛化性能,更适合对分散程度较大、聚类性较差的航空发动机故障样本进行分类。     

基于CORBA实现专业领域软件在网络环境下的集成

如何有效地将已有系统集成或移植到新系统中 ,是集成系统开发中提出的一个重要问题。本文基于CORBA技术探讨如何实现遗留系统集成的方法 ,并通过实例重点阐述如何将已有的系统封装成CORBA对象的步骤。     

加劲板的二维复合型广义J积分

 本文以虚功原理与本构方程为基础导出了加劲板二维复合广义J积分J₁与J₂,证明了它们的收敛性、守恒性并且导出了J₁、J₂与应力强度因子K₁K₁的关系。在此基础上,本文采用两种有限元计算了含有斜裂纹的加劲板应力强度因子。计算结果表明,这个复合广义J积分的守恒性很好,由两种元素所得结果相当一致,相对误差在工程所允许的范围之内。     

大上翘角机身后体流动机理研究

通过对圆截面上翘 1 6°的后体的数值计算 ,研究了上翘后体的流动机理。结果表明 ,上翘后体引起横向流动 ,使横向逆压梯度增大、下表面边界层增厚 ,导致后体出现三维开式分离流动 ;由分离形成一对方向相反、强度相等的旋涡向下游拖至尾迹区 ;这种分离流动是导致大上翘角后体阻力增加的主要原因。     

凹槽火焰稳定器结构对煤油超声速燃烧的影响

采用混合分数平衡化学反应模型和离散液滴模型，研究了不同长深比和后缘倾角的凹槽在液体煤油双模态燃烧中的火焰稳定特性。结果表明，集燃料喷射、混合及火焰稳定为一体的凹槽，在所模拟的工作过程中增强了燃料与空气的混合和燃烧；并得到最优的凹槽结构；混合分数平衡化学反应模型和离散液滴模型能准确地预测煤油在双模态燃烧室内的喷雾燃烧。     

用于棘轮变形预测的棘轮演化统一模型研究

 在大量单轴棘轮实验基础上,研究了均值、幅值、峰值和谷值应力对 304不锈钢的饱和棘轮应变的影响规律,首次提出了棘轮应力σ_r和棘轮门槛值σ_rth 的概念,建立了基于单参数控制的、用于饱和棘轮应变预测的 SRM饱和棘轮本构模型和用于独立循环应力工况下棘轮应变演化预测的 REM棘轮演化模型,并由此发展了全面描述任意循环应力工况下棘轮应变演化规律的 URM棘轮演化统一模型。抛物律模型 SRM和幂律模型 REM对实验数据拟合的相关系数均超过 0.98。URM建模容易,只需 4～ 6个试样的单轴棘轮实验数据。此外还讨论了获得 SRM本构模型的单试样实验法。     

文化制胜对企业竞争力的提升

企业文化的建立在企业发展中显示出越来越重要的作用。从文化制胜的内涵出发,分析了企业文化对企业竞争力提升的作用,以及如何实现企业文化制胜。     

基于知网的汉语语义实例库的建设与应用

为了解决有指导词义消歧任务中的知识源问题,提出并构建了一个基于知网的汉语词义实例库(CSIC),同时开发了一个语义标注平台(SenseTag)。该平台通过方便快捷的人机交互方式显著提高了实例库的建设效率和质量。利用条件随机场模型,从实例库中自动学习消歧知识,进行自动标准在随机选取的部分汉语高频多义词的词义消歧开放测试中,取得了平均正确率85%的较好效果。     

Expectations for Infrared Spectroscopy of 9P/Tempel 1 from Deep Impact

The science payload on the Deep Impact mission includes a 1.05–4.8 μm infrared spectrometer with a spectral resolution ranging from R∼200–900. The Deep Impact IR spectrometer was designed to optimize, within engineering and cost constraints, observations of the dust, gas, and nucleus of 9P/Tempel 1. The wavelength range includes absorption and emission features from ices, silicates, organics, and many gases that are known to be, or anticipated to be, present on comets. The expected data will provide measurements at previously unseen spatial resolution before, during, and after our cratering experiment at the comet 9P/Tempel 1. This article explores the unique aspects of the Deep Impact IR spectrometer experiment, presents a range of expectations for spectral data of 9P/Tempel 1, and summarizes the specific science objectives at each phase of the mission.     

Deep Impact: Working Properties for the Target Nucleus – Comet 9P/Tempel 1

In 1998, Comet 9P/Tempel 1 was chosen as the target of the Deep Impact mission (A’Hearn, M. F., Belton, M. J. S., and Delamere, A., Space Sci. Rev., 2005) even though very little was known about its physical properties. Efforts were immediately begun to improve this situation by the Deep Impact Science Team leading to the founding of a worldwide observing campaign (Meech et al., Space Sci. Rev., 2005a). This campaign has already produced a great deal of information on the global properties of the comet’s nucleus (summarized in Table I) that is vital to the planning and the assessment of the chances of success at the impact and encounter. Since the mission was begun the successful encounters of the Deep Space 1 spacecraft at Comet 19P/Borrelly and the Stardust spacecraft at Comet 81P/Wild 2 have occurred yielding new information on the state of the nuclei of these two comets. This information, together with earlier results on the nucleus of comet 1P/Halley from the European Space Agency’s Giotto, the Soviet Vega mission, and various ground-based observational and theoretical studies, is used as a basis for conjectures on the morphological, geological, mechanical, and compositional properties of the surface and subsurface that Deep Impact may find at 9P/Tempel 1. We adopt the following working values (circa December 2004) for the nucleus parameters of prime importance to Deep Impact as follows: mean effective radius = 3.25± 0.2 km, shape – irregular triaxial ellipsoid with a/b = 3.2± 0.4 and overall dimensions of ∼14.4 × 4.4 × 4.4 km, principal axis rotation with period = 41.85± 0.1 hr, pole directions (RA, Dec, J2000) = 46± 10, 73± 10 deg (Pole 1) or 287± 14, 16.5± 10 deg (Pole 2) (the two poles are photometrically, but not geometrically, equivalent), Kron-Cousins (V-R) color = 0.56± 0.02, V-band geometric albedo = 0.04± 0.01, R-band geometric albedo = 0.05± 0.01, R-band H(1,1,0) = 14.441± 0.067, and mass ∼7×10¹³ kg assuming a bulk density of 500 kg m⁻³. As these are working values, {i.e.}, based on preliminary analyses, it is expected that adjustments to their values may be made before encounter as improved estimates become available through further analysis of the large database being made available by the Deep Impact observing campaign. Given the parameters listed above the impact will occur in an environment where the local gravity is estimated at 0.027–0.04 cm s⁻² and the escape velocity between 1.4 and 2 m s⁻¹. For both of the rotation poles found here, the Deep Impact spacecraft on approach to encounter will find the rotation axis close to the plane of the sky (aspect angles 82.2 and 69.7 deg. for pole 1 and 2, respectively). However, until the rotation period estimate is substantially improved, it will remain uncertain whether the impactor will collide with the broadside or the ends of the nucleus.