基于神经网络的某鱼雷保障设备故障预测方法

为确保鱼雷保障设备处于良好的工作状态,提出一种基于径向基函数(radial basic function,RBF)神经网络的某鱼雷保障设备故障预测法。利用RBF神经网络的非线性建模能力,在某鱼雷保障设备的关键监测点建立网络诊断模型,通过对该模型的训练学习,确定需要的参数估计,再根据该模型的输出值来判断故障,并在Matlab仿真环境下对该设备故障进行了预测,其预测结果与实际情况基本一致。仿真结果表明:RBF神经网络作为预测网络能较好地解决该保障设备的故障预测问题,具有较准确和快速的诊断能力,可为复杂设备的预防性维修提供科学依据。     

基于改进深度置信网络的故障诊断方法

针对深度置信网络对小样本数据集故障诊断时准确率低的问题,提出了基于改进深度置信网络的故障诊断方法。该方法通过优化网络特征提取能力,提升网络学习和分类能力以减少网络训练对数据的依赖程度;利用网络公开数据集测试改进的深度置信网络模型的性能;并将改进网络模型应用于小样本故障数据集上。实验验证结果表明,较基于传统网络模型的故障诊断而言,基于改进模型的方法通过添加新的隐藏层,强化了模型特征提取能力,提高了故障诊断准确率。     

基于深度置信网络的通信控制设备故障诊断

针对传统的故障诊断方法在装备故障诊断中带来的准确率低问题,提出了一种基于小波变换(WT)和深度置信网络(DBN)的通信控制设备故障诊断方法。该方法通过采集样本故障数据、分析处理故障数据、输入神经网络进行学习、对测试数据进行识别等步骤对通信控制设备进行故障诊断,并与支持向量机(SVM)、k类临近法(KNN)和BP神经网络等传统故障诊断方法进行比较。仿真实验结果表明,提出的方案识别准确率达到93%,与传统方法相比具有更高的准确率,表现更好。     

面向发动机故障诊断精度的深度随机森林优化研究

为解决装备故障检测存在故障数据少,难以应用深度学习方法提高性能的问题,提出一种基于优化深度 随机森林的提高装备故障诊断精度方法。根据数据集合数据的特点,构建重采样的样本集合决策树模型,通过对故 障数据中的连续数值进行C4.5 决策树离散化处理,使用扩张卷积方法扩展滑动窗口的感测范围,对训练模型进行实 验验证。实验结果表明,运用深度随机森林的方法对提高装备故障诊断有一定借鉴作用。     

基于长短时记忆网络的仿真系统数据故障诊断方法

针对空空导弹仿真系统中传统数据分析方法效率低下,无法快速有效对大量仿真数据进行故障识别的问题,提出基于长短时记忆网络的仿真系统数据故障诊断方法。该方法分析仿真试验数据的数据特征,采用长短时记忆网络对健康数据中的时序信息进行记忆,并内嵌至所建立的多层次深度神经网络模型,从而有效表征健康数据的本征变化属性,进而对试验过程中的故障数据进行准确诊断。空空导弹仿真系统数据智能诊断结果表明,本方法在快速诊断的基础上,故障诊断识别率达到90%以上。     

基于BP神经网络的导弹故障诊断专家系统设计

分析了某型导弹测试系统的故障机理,针对这种故障机理,结合故障诊断需求,设计了故障诊断专家系统,建立了BP神经网络模型,并利用训练样本对网络进行了训练,重点针对导弹模飞测试故障诊断模块进行了应用分析．解决了故障分析、判断不够精确、自学习能力不强等问题,为该型导弹测试系统的故障智能诊断提供了一种新的实现方法．     

神经网络专家系统在发动机故障诊断中的应用

将神经网络与专家系统相结合形成了神经网络专家系统,其核心部分采用BP神经网络．以发动机为诊断对象,获取故障数据后,采用3层BP神经网络模型对网络进行训练,经过几种不同网络参数的对比后,设定了合理的网络训练参数,训练后获得了期望的误差值0．01,并采用VB6．0软件设计了专家系统的人机交互界面,为专家系统在发动机故障诊断中的应用提供了有益的参考．     

基于故障树与灰关联分析的鱼雷故障诊断

将灰色关联度与故障树联合运用于鱼雷故障诊断,为鱼雷的故障诊断提供了一种新途径。在鱼雷动力系统故障树诊断模式识别中,运用灰色系统理论中的灰关联分析,通过进行关联度计算及排序,对各种故障模式发生的可能性大小作出了判断,从而为处理故障的轻重缓急、控制故障的发生以及改进系统的可靠性提供了理论依据。实例诊断表明,这种方法是有效可行的,并且诊断结果可靠。     

基于改进BP神经网络的模拟电路故障诊断研究

根据BP神经网络特点,提出对BP神经网络改进的方法,并以某负反馈放大器为例,采用改进后的BP网络进行故障诊断,步骤包括：故障特征向量提取、原始数据归一化处理、BP网络设计与训练。结果表明,在MATLAB7．1中运用神经网络工具箱中函数进行仿真,能有效进行故障识别、改善神经网络结构、提高故障诊断精度和速度。     

基于卷积神经网络的滚动轴承故障诊断

针对传统故障诊断方法诊断过程复杂、效果不佳的问题,提出一种基于卷积神经网络的滚动轴承故障诊 断方法。首先选取不同故障的振动信号进行归一化处理,然后把 1 维的振动信号转化成 2 维的灰度图像,利用每个 元素与其相邻元素之间的关系,并且采用重叠采样的方法加强数据集。在卷积神经网方面利用 tensorflow 搭建网络 框架,采用 4 种不同的卷积神经网络结构对样本进行训练。为避免实验的随机性,对每种方案进行多次训练,采其 结果的均值。根据测试集的准确率选取最好的适合轴承故障诊断的模型,同时对网络的结构参数进行优化改进,提 高模型的识别率和运行效率。实验结果表明,该方法可以准确地将滚动轴承的故障进行识别和分类。     

Nitrosation and Nitration of Tetraacetyldibenzylhexaazaisowurtzitane

Tetraacetyldibenzylhitane (TADBIW) was subjected to debenzylation by nitrosating with inorganic materials available commercially to synthesize tetraacetyldinitrosohexaazaisowurtzitane (TADNSIW). TADNSIW was purified, and its structure was determined by FTIR, 1H NMR, MS and element analysis. The debenzylation reaction of TADBIW gave quantitative benzaldehyde as a by-product. This indicates that the reaction produces an imine cation as an intermediate. Hexanitrohexaazaisowurtzitane (HNIW) was prepared from unpurified TADNSIW with the yield over 96.0 % and the purity more than 98.0 %. And the mechanism of the reaction from TADNSIW to HNIW is proposed to be oxidation of nitroso and nitration of acetyl on the molecule of TADNSIW, This reaction system involved is simple, and the reaction can complete within a short time and under mild conditions. The product can be easily to separate and the waste disposed readily.     

A Robust Algorithm for Watermarking 2D Vector Maps with Low Shape-Distortions

The technique of watermarking has been introduced into 2D vector maps for many potential applications such as protecting the copyright of important GIS （geographical information system） data and tracing the data source of military maps. In most previous works, although the validity of the map data could be preserved during the embedding procedure, the shape-distortions of the map elements are usually neglected which would tend to degrade the invisibility of the watermarking schemes. A shape-preserving algorithm for watermarking 2D vector maps is presented in this paper. A 1D distance sequence extracted from the original map is adopted as the cover data instead of 2D coordinates. A watermark bit is represented by changing the distributions of the cover data. The embedding procedure induces lower shape-distortions than in former works. Experimental results indicate better invisibility of the proposed scheme, as well as its robustness to certain attacks such as map simplification, interpolation, additive noise and most geometric transforms.     

Open Rule-setting System Applied to Conceptual Design of Engines

Open rule-setting method advanced in this paper concentrates on providing designers of engines with a convenient way to express their design innovations and develop the expected prototypes in the early CAD stage, and constitute effective models for the following analysis process of CAE and CAPP. The problems arisen in the process of conceptual design with the traditional experience-based development method are analyzed. Based on those analyses, open rule-setting method is presented and some associated technical problems are discussed. The functional framework of open rule-setting system was built as software engineering methods. The speciality of the engine product as a kind of complex product and the requirement of the engine product based on its structural particularity to rule-setting system are considered carefully. A demonstration is supplied to illustrate how the open rule-setting method enhances the efficiency and quality of the engine conceptual design.     

Research on Linux Network Device Driver

By analyzing the working principle of Linux network device driver, discussing the Linux network driver structure and its key technologies, the general network driver structure and its design methodologies in Linux system are summarized. Through modifying the network device driver of Linux device driver 3rd version snull, c and improving the COW technology, the Zero-Copy technology in Linux （kernel version 2.6.11 ） is implemented. In the end, the success test tells us that the thorough analysis of network device driver is the foundation of many applications, and it also provides a certain improvement to a lot of real applications, even to military application development.     

Method for Measuring Velocity of Warhead Fragments Based on Photoelectric Detection

For measuring velocity and impacting position of single fragment of warhead, a non-contact measuring method is proposed, in which a six-light-screen array, a position indicator, a multi-channel chronograph and a computer are used.The principle of measurement is described. The key device of the system is a light screen array sensor which consists of six light screens allocated with certain geometrical parameters. When the fragment flies through the light screen array, the time of passing through each of the screens is recorded by the multi-channel chronograph. According to the time data and the geometrical parameters of the array, the velocity vector and the location of the fragment can be calculated immediately. The presented method can be used to locate the fragment and to measure the real velocity on its flying direction. It can also be used to measure the velocity of a fragment swarm after the system is engineered further.     

数码电子雷管中影响点火头性能因素分析

阐述了数码电子雷管中,点火药剂细结晶三硝基间苯二酚铅（LTNR）的制备工艺;从点火药剂的种类、加入黏合剂的质量分数、桥丝直径、点火头的电阻值、防潮漆的蘸涂等方面对点火头性能的影响因素进行了探讨;通过震动试验、并联起爆试验、延期精度检测与考核,试验结果表明：一次性合成的细结晶LTNR是制造数码电子雷管点火头较为理想的点火药剂,按技术要求生产的点火头性能指标能够满足行业标准要求。     

Improvement and Experiment of the Reamer in Vibra Cutter Against Wave Edge of Titanium Sheet Alloy Parts

By analyzing the influencing factors of part quality making of sheet alloy of titanium by vibra cutter, the shape of upper reamer is set as cylinder and wedge-shaped form, and the lower reamer as plain and hemicycle form, and its main structural parameters are defined as well. Then it is validated further that such improved vibra cutter reamer can be used to process curve-edged parts of titanium alloy sheet. The experimental result shows that the titanium alloy sheet parts processed by above equipment have no sharpen angles for convex parts and evident crevasse of concave-edged part. In summary, such improvement can eliminate the free-waved edge and improve the manufacture quality of titanium alloy sheet parts greatly.     

Experimental Study on the Small Flyer Initiating Behavior of Fine-grained Explosives

The initiating behavior of fine-grained explosives by small flyer is studied. The diameter of small flyer in this device is 1 ram. The test results indicate that the granularity of explosives has great effect on its flyer initiating sensitivity.The flyer initiating sensitivity of the fine-grained explosives is higher and the critical initiating energy is lower than that of common explosives. For common explosive, the flyer initiating sensitivity increases as the density is reduced. But for the fine-grained explosive, the test results are exactly opposite.     

Algorithm for Multi-laser-target Tracking Based on Clustering Fusion

Multi-laser-target tracking is an important subject in the field of signal processing of laser warners. A clustering method is applied to the measurement of laser warner, and the space-time fusion for measurements in the same cluster is accomplished. Real-time tracking of multi-laser-target and real-time picking of multi-laser-signal are introduced using data fusion of the measurements. A prototype device of the algorithm is built up. The results of experiments show that the algorithm is very effective.     

Influence of Installing Error on the Surface Shape Precision of the Axial Symmetry Aspheric Workpiece in Lapping

A method to calculate the surface shape error, which is caused by the installing error between the workpiece and the lapping tool in the process of form lapping, is proposed. The mathematical model which the installing translation error influences on the workpiece surface shape error is established. The changing rule of the error is simulated through the calculating example of the paraboloid workpiece. The results indicate that the surface shape error of the workpiece is increasing with the increase of the installing translation error, it is also increasing gradually along the center point of the curve surface to the edge, and the influence is severer to the curve surface with great curvature than that of the small curvature when the translation error is the same.