基于核方法的模拟电路故障诊断

核方法通过非线性映射将原始数据嵌入到高维特征空间,然后进行线性分析和处理,为基于知识的数据分析带来新的方法和模式;传统方法无法解决故障特征数据维数高、在故障样本交叠严重时多分类性能较差的问题,因此在电路故障特征数据预处理阶段,提出了分步骤分别在时域对电路输出电压波形进行小波包分析和在频域测量电路幅频特性的方法来提取电路故障特征;预处理后的故障特征向量只是8维向量,减少了SVM的训练时间;将该方法应用于国际标准电路中的CTSV滤波器电路的故障诊断,结果表明:该方法能突出不同故障的特性,故障诊断正确率达到98.57%(414/420)。     

基于盲反卷积去噪的机车走行部齿轮箱故障诊断

为了有效地识别机车走行部的早期故障,提高我国重载机车的运输能力,以机车走行部齿轮箱为例,在分析总结大量实际振动信号的基础上,探讨了运用盲反卷积去噪的方法来提取机车齿轮的故障特征信息,即采用傅里叶变换在送入滤波器中使振动信号变换为时域信号;再对所有的盲信号进行盲源分离,与TSA(时间同步平均)方法相比,考虑了外界因素的影响,尽量避免了预测和诊断误差;通过对实际机车走行部齿轮振动信号的分析,成功提取到故障特征信号,阐述了该方法在机车走行部在线故障诊断系统中的应用,不仅能满足在线监测和故障的实时性和可靠性的要求,使机车在正常运行的时候,同时显示出其可靠性、耐用性和高置信度性。     

导弹逼近告警器及无源干扰设备测试软件设计

为解决导弹逼近告警器及无源干扰设备测试成本高、维护保障难的问题,研制了一套导弹逼近告警器及无源干扰设备自动检测系统;设计了系统测试软件,按照模块化设计方式,针对设备检测需求,开发不同的软件测试模块并集成;该系统集成度高,配置灵活,操作智能化,功能完整,测试项目全,测试和排故障效率高;经试用表明,该检测系统具有良好的操作性和维护性,可以为导弹逼近告警器及无源干扰设备检测维修提供技术支持,有效节省测试成本和维护保障费用。     

基于直推式学习的网络故障诊断算法

网络故障的及时诊断能够保证日常工作、学习和生活能够正常进行。传统的基于监督式学习的诊断方法依赖于大量具有鉴别意义的样本,这在实际情况中通常难以得到满足。针对上述问题,本文提出了一种基于直推式学习的诊断算法。针对大规模的网络管理的特征数据,本算法利用主成分分析对特征进行降维,并利用新的度量下的特征数据来构建拉普拉斯矩阵;该矩阵能够很好的描述带检测样本和训练样本之间的关系。在此基础上,本文设计了基于直推式学习的目标函数,并利用拉格朗日乘子法完成了优化。实验部分证明了本算法能够在有限数目的带标签的样本的前提下获得精确的分类结果,能够显著提高网络故障诊断的检测率。     

时域同步平均技术在直升机主减速器故障诊断中的应用分析

由于直升机主减速器的结构特点致使振动信号的信噪比很低,给故障特征的提取带来了难度;文中以某型直升机的主减速器为研究对象,开展时域同步平均技术在主减速器故障特征提取上应用的深入研究,分析了该技术在实际中的应用以及在应用过程中的误差影响,在此基础上提出了时域同步平均误差分析的改进方法,通过在直升机主减速器飞行数据的对比应用分析,证明该方法的可行性和有效性,并能在一定程度上提高故障特征提取的精度。     

基于迭代学习的离散切换系统故障估计

针对一类具有任意切换序列的离散切换系统的故障估计问题,提出了一种新的故障估计算法.该算法利用引入的虚拟故障信号构建出故障估计器,并利用残差信号通过迭代学习方法对引入的虚拟故障进行逐次修正.使虚拟故障随着迭代次数的增加逐渐逼近实际故障.利用压缩映射方法严格证明了算法在各个子区间上的收敛性,给出了算法的收敛条件.理论分析表明,所提算法能够在有限区间上精确估计出切换系统发生的不同类型故障.最后通过仿真实验进一步验证了所提算法的有效性.     

指向Lyapunov指数及其在单输入单输出系统故障检测中的应用

针对单输入单输出系统的故障检测, 采用混沌振荡器作为激励源, 并利用非一致延迟时间法对被测系统输出时间序列进行相空间重构. 在相空间中平衡点附近定义了指向Lyapunov指数, 并用其对被测系统输出在相空间中平衡点附近特征结构进行分析, 实现了对单输入单输出系统的故障检测. 仿真结果表明, 被测系统的参数变化将会引起相空间中平衡点附近特征结构的改变, 指向Lyapunov指数对其变化敏感. 关键词： 混沌激励 指向Lyapunov指数 故障检测 单输入单输出系统     

Comparison of formation and evolution of radiation-induced defects in pure Ni and Ni-Co-Fe medium-entropy alloy

High-entropy alloys (HEAs) and medium-entropy alloys (MEAs) have attracted a great deal of attention for developing nuclear materials because of their excellent irradiation tolerance. Herein, formation and evolution of radiation-induced defects in NiCoFe MEA and pure Ni are investigated and compared using molecular dynamics simulation. It is observed that the defect recombination rate of ternary NiCoFe MEA is higher than that of pure Ni, which is mainly because, in the process of cascade collision, the energy dissipated through atom displacement decreases with increasing the chemical disorder. Consequently, the heat peak phase lasts longer, and the recombination time of the radiation defects (interstitial atoms and vacancies) is likewise longer, with fewer deleterious defects. Moreover, by studying the formation and evolution of dislocation loops in Ni-Co-Fe alloys and Ni, it is found that the stacking fault energy in Ni-Co-Fe decreases as the elemental composition increases, facilitating the formation of ideal stacking fault tetrahedron structures. Hence, these findings shed new light on studying the formation and evolution of radiation-induced defects in MEAs.     

Evolution of past enamel technology and metal conservation issues: the case of two Byzantine style bindings

The results of a Raman spectroscopic study of the cloisonne’ and basse‐taille enamels, which beautify two Byzantine style bindings from the Marciana Library (Venice), namely the Lat. III,111 and the MsGR.I.53 codexes, are presented in this work. The first binding dates back to the 13th century and was subject to an early restoration work in the 14th century, when new enamels substituted four originals. The second binding, from the 15th century, shows a lower number of enamels, all originals, and with a larger color palette. The white and yellow enamels of both codexes were successfully characterized and the red ones, where hematite was not used. Interestingly the white and yellow color of the 13th century enamels of the Lat. III,111 codex has been obtained by an ancient technique of the glass technology, which was already obsolete in the 13th century, and is based on the use of calcium antimonate and Naples yellow. The white color in the other binding's enamels have been instead obtained by using cassiterite, according to the tradition of the time. Cassiterite was also mixed to Naples yellow in the enamels of the MsGR.I.53 codex, to change the yellow hue. The identification of agents determining some colors is instead uncertain or not feasible by Raman spectroscopy. The transparency of the 14th century enamels of the Lat. III,111 codex has allowed the comparison of metal degradation below the enamels and in regions exposed to the atmosphere. Copyright © 2012 John Wiley & Sons, Ltd.     

基于多种群混沌遗传神经网络的模拟电路故障诊断

针对标准遗传算法优化BP神经网络收敛慢,易陷入局部最优的问题,提出了改进的多种群协同进化遗传算法,该算法改变了以往的随机初始化方法,采用了附加混沌扰动的tent映射初始化均匀分布的种群,提高了初始解的质量;每个种群采用自适应交叉率和变异率,引入移民算子实现种群间的横向联系;算法通过多种群的协同进化和种群间的个体移植提高了算法的搜索均匀性和效率;仿真实验表明该算法误差小,收敛速度快,诊断正确率高,较好地解决了模拟电路的软故障诊断问题。