基于磁致伸缩作动器的主动隔振系统分析

基于材料的本构方程,建立主动隔振系统的模型,并对其动力学和控制稳定性进行了研究。分析和仿真表明,上层质量对主动控制力的影响较大,采用中间质量的加速度作为反馈量时下层刚度对隔振效果影响明显。在系统的前馈控制形式中,振动频率的变化不会影响系统稳定性,但反馈环节将会降低稳定裕度;在反馈控制形式中,随着振动频率的增大,次通道的延迟作用可能使系统不稳定。      

A theoretical model for the magnetic helicity of solar active regions

Active regions on the solar surface are known to possess magnetic helicity, which is predominantly negative in the northern hemisphere and positive in the southern hemisphere. Choudhuri et al. [Choudhuri, A.R. On the connection between mean field dynamo theory and flux tubes. Solar Phys. 215, 31–55, 2003] proposed that the magnetic helicity arises due to the wrapping up of the poloidal field of the convection zone around rising flux tubes which form active regions. Choudhuri [Choudhuri, A.R., Chatterjee, P., Nandy, D. Helicity of solar active regions from a dynamo model. ApJ 615, L57–L60, 2004] used this idea to calculate magnetic helicity from their solar dynamo model. Apart from getting broad agreements with observational data, they also predict that the hemispheric helicity rule may be violated at the beginning of a solar cycle. Chatterjee et al. [Chatterjee, P., Choudhuri, A.R., Petrovay, K. Development of twist in an emerging magnetic flux tube by poloidal field accretion. A&A 449, 781–789, 2006] study the penetration of the wrapped poloidal field into the rising flux tube due to turbulent diffusion using a simple 1-d model. They find that the extent of penetration of the wrapped field will depend on how weak the magnetic field inside the rising flux tube becomes before its emergence. They conclude that more detailed observational data will throw light on the physical conditions of flux tubes just before their emergence to the photosphere.     

X-ray study of accretion flow in narrow-line Seyfert 1 galaxies

We present the results of a systematic study of narrow-line Seyfert 1 galaxies (NLS1s) observed with XMM-Newton. The 2–12 keV X-ray spectra of NLS1s are well represented by a single power law with a photon index Γ ∼ 2. When this hard power law continuum is extrapolated into the low energy band, we found that all objects in our sample show prominent soft excess emission. This excess emission is well parameterized by the thermal emission expected from an optically thick accretion disk, and we found the following three peculiar features: (1) The derived disk temperatures are significantly higher than the expectation from a standard Shakura-Sunyaev accretion disk, if we assume a central mass of a black hole to be 10^6–8M_⊙. (2) The temperatures are distributed within narrow range (ΔkT ∼ 0.08 keV) with an average temperature of 0.18 keV in spite of the range of four orders of magnitude in luminosity (10^41–45 erg s⁻¹). (3) We found a peculiar temperature–luminosity relation, where the luminosity seems to be almost saturated in spite of the significant change in temperature, during the observations of the most luminous NLS1 PKS 0558-504. These results strongly suggest that the standard accretion disk picture is no longer appropriate in the nuclei of NLS1s. We discuss a possible origin for the soft excess component, and suggest that a slim disk may be able to explain the observational results, if the photon trapping effect is properly taken into account.     

半埋入式S弯进气道主动流动控制研究

为了改善半埋入式亚声速S弯进气道出口气流品质,本文采用吹、吸气方式对其进行了主动流动控制数值研究。结果表明:吸气位置、吸气量变化显著影响进气道气动性能,最佳吸气位置位于唇口附近,当唇口吸气量为进气道流量的1.9%时,总压畸变指数降低约45.7%,总压恢复系数和增压比略有提高;而附面层吹气时进气道的性能仅在吹气量较小时有所改善;附面层吹、吸组合控制对流场的改善程度介于上述方案之间,而非单独吹、吸气方案效果的简单叠加。     

基于深度学习的雷达辐射源识别算法

为解决雷达辐射源识别中特征提取困难、低信噪比条件下识别效率低的问题,提出了一种基于一维卷积神经网络和长短期记忆网络的深度学习智能识别算法,构建了一个CNN-LSTM网络,能实现对不同脉内调制方式雷达辐射源的端到端识别.该网络首先利用卷积层学习信号局部特征,然后将卷积层输出的结果输入长短期记忆网络,学习信号的全局特征,最...     

人工智能可解释性评估研究综述CSCD

近年来,可解释人工智能(XAI)发展迅速,成为当前人工智能领域的研究热点,已出现多种人工智能解释方法。如何量化评估XAI的可解释性以及解释方法的效果,对研究XAI具有重要意义。XAI的可解释性评估涉及主、客观因素,是一个复杂且有挑战性的工作。综述了XAI的可解释性评估方法,首先,介绍了XAI的可解释性及其评估的概念和分类;其次,总结和梳理了一些可解释性的特性;在此基础上,从可解释性评估方法和可解释性评估框架两方面,综述和分析了当前可解释性评估工作;最后,总结了当前人工智能可解释性评估研究的不足,并展望了其未来发展方向。     

Active detumbling technology for noncooperative space target with energy dissipation

Space debris is generally a kind of tumbling noncooperative space target which poses a serious threat to human space activities. In active debris removal (ADR) missions, capturing a space target directly may cause damage to space manipulator and chaser satellite, so it is a feasible strategy to reduce the angular velocity of space target to an acceptable range in the pre-capture phase. In this paper, an active detumbling technology for a free-floating tumbling space target with energy dissipation is studied, and an effective detumbling method utilizing intermittent contact impact between the space target and despin mechanism is proposed. First, the dynamic model of the space target is set up by the Jourdain’s velocity variation principle. Then, a contact model between the space target and despin mechanism is established based on the Hertz contact theory and the method of computer graphics. Finally, the detumbling method is validated by numerical simulations. Simulation results show that our method can reduce the angular velocity of the space target effectively without causing large nutation.     

激光选区熔化增材制造缺陷智能监测与过程控制综述

激光选区熔化（SLM）技术被认为是最有应用前景的增材制造技术之一,已应用于航空航天、医疗器械等领域。然而,如何确保构件质量的可靠性和制造的可重复性是SLM面临的最大挑战,已被认为是限制SLM及其他金属增材制造技术发展和工业应用的最大壁垒。其中,主要原因是SLM过程中会产生难以控制的缺陷。因此,对SLM进行过程监测和实时反馈控制是解决这一挑战的重要研究方向,也已成为学术界和工业界的研究热点之一。通过对近十年该领域的文献调研,综述了金属激光增材制造中常见的冶金缺陷及其产生机理,对金属增材制造过程产生的信号及其监测手段,如声信号、光信号及热信号等进行了详细描述;总结了信号数据的处理方法,包括传统的统计处理方法和新兴的基于机器学习的智能监测方法;随后,综述了金属增材制造过程的质量控制方法,包括非闭环控制和闭环控制,并对全文进行了总结,展望了未来SLM智能监测和控制领域值得深入的研究方向。     

超燃冲压发动机燃烧流场纹影图像无监督分割方法

超燃冲压发动机燃烧室流场纹影图像常存在大量噪声信号,如何高效、准确提取燃烧流场图像的主要波系结构成为当前亟需探索的问题。以超燃冲压发动机燃烧室内冷流到氢燃料点火阶段流场为研究对象,基于深度神经网络方法,发展一种燃烧室内流场的关键波系结构快速识别方法。首先,采用基于图论的超像素分割方法对纹影图像进行聚类分割,为语义信息明显相同区域分配伪标签;其次,设计了一种全卷积特征提取神经网络,并使用残差结构对各个通道进行加权,提取纹影图像高级语义特征;最后,使用交叉熵目标函数优化网络模型,并通过阈值滤波操作去除噪声像素点,提升语义分割效果。结果表明：与K-means及自适应高斯阈值方法相比,本文提出方法在准确率、召回率、F1分数和交并比指标性能明显是最优的,能够准确完成燃烧流场纹影图像语义分割任务,可以更加清晰地反应流场内的主要波系和剪切层结构     

情报侦察数据融合技术的几个问题

就情报侦察系统中多传感器 (有源 -无源 )数据融合技术的坐标变换、目标状态估计特征、数据关联和跟踪维持、目标身份识别等问题进行了讨论 ,并就各种相关准则和算法给予研究 ,特别提出了模糊聚类关联算法。