基于Shannon熵的飞机交流发电机振动分析

飞机交流发电机是飞机关键部件之一,监测其健康状况具有重要意义。通过设计实验数据获取平台采集交流发电机在转速上升(从6000转/分到8500转/分)、平稳运行(7500转/分)及转速下降(8500转/分到6000转/分)三个阶段的振动信号,利用符号序列Shannon熵的计算方法以及符号化方法设定合适参数对振动信号进行了处理,得出了交流发电机在三个阶段的振动信号的Shannon熵;通过初步的分析表明,交流发电机在正常状态下三个阶段的Shannon熵都比较有规律的在一定范围内变化,一旦发生振动异常将会引起Shannon熵突变;经过分析讨论,Shannon熵的突变与否是可以用来反映交流发电机在机械方面是否存在故障隐患的。     

基于北斗卫星的森林火灾监控终端设计

针对传统森林火灾监控方式存在的受地形限制,导致火灾发现不及时和定位难的缺陷,设计了基于北斗卫星的森林火灾监控系统,实现了火灾智能识别和预警,并且可以对火源、火场进行快速定位和信息通报;终端包括烟采集处理模块、图像采集模块、DSP处理模块、ARM主控模块和北斗卫星通信模块五部分,从烟和火焰两个方面对森林实时进行监控,分别采用离子烟雾传感器和DSP火焰识别混合算法,提取火焰图像差值和火焰图像边界参数的方法,提高了识别的准确率;试验证明,系统具有结构简单、使用方便和识别智能化的特点。     

气动人工肌肉手臂的神经网络Smith预估控制

针对含时滞d的1关节气动人工肌肉(PAM)手臂,用三层递归神经网络(RNN),建立PAM手臂包含时滞的模型(即非线性Smith预估器),并超前d步预测PAM手臂的输出角度.将此超前d步的预测值作为反馈量,与设定值相比较得到的误差作为PID控制器输入量,实现Smith预估PID控制.同时每一步都用RNN模型当前时刻的输出值与PAM手臂当前时刻实际输出值之差的平方做为RNN权值的在线调整准则对RNN预测模型的权值进行在线调整,以自适应PAM手臂的不确定性和时变性.使用Matlab通过串口和研华亚当模块对实物PAM手臂进行控制,控制效果表明所提出的Smith预估PID控制算法比常规PID控制算法的性能有显著提高,证明所提出的算法是有效的和切实可行的.     

基于ARM的嵌入式语音存储系统设计

为了有效地节省语音数据的传输带宽和存储系统的磁盘空间,需要在保证语音质量的前提下尽可能降低其编码比特率。本设计采用经过优化的G．729语音压缩编译码算法,以ARM处理器为载体,开发的嵌入式语音存储系统可实现语音信号的海量存储,而且处理速度快、可靠性好、扩展方便。通过严格的测试和评估,该系统能够实现对大量语音数据的压缩和记录,各项指标基本达到了预期的水平。     

Developing feasible and optimal schedules of jobs on one machine

The paper considers the properties of feasible and optimal scheduling of jobs on one machine under constraints on the terms of the beginning and completion of jobs and on partial sequences of job performance. The established properties and the lower-bound estimates of the length of the optimal schedule are used to develop methods for the exact and approximate solutions of the formulated problem by sequential optimization algorithms. The proposed algorithms are illustrated by numerical examples and can be successfully applied to solve these problems in the absence of constraints.     

Comparative evaluation of linear edge detection methods

A comparative evaluation of the most commonly used linear methods for edge detection in grayscale images are presented. Detectors based on the first and second derivatives of image brightness are considered. The method for automatic edge tracking in grayscale images is proposed. The model for assessing errors and artifacts caused by sampling during digitization of real input images is proposed. Investigation of edge detectors isotropy and errors caused by input images sampling is conducted. The advantage of the Isotropic operator for edge tracking is shown. The noise immunity of linear edge detection methods is assessed and the superiority of 3 × 3 gradient operators for noisy images is shown. Isotropic and Sobel operators are identified to be optimal on a basis of sampling errors, output noise level, and computational complexity.     

Control of downrange and crossrange motion of a descent module in a certain range of reentry angles with load factor constraints

Results concerning the terminal guidance algorithm at the reentry of the descent module into the Earth atmosphere are presented. This algorithm simultaneously eliminates predicted deviations in the downrange and crossrange directions. The restriction on the tolerable load factor n ?? 3 is taken into account. Three reentry angles are analyzed: the medium angle ?1.5°, the steep angle ?1.9°, and the flat angle ?1.2°. The motion of the descent module is considered in the disturbed Earth atmosphere represented by the TsNIIMash model. This model includes variations of the atmosphere density and the wind field. It is demonstrated that, depending on the reentry angle, the proposed algorithm guarantees the downrange maneuver zone in the range 1240?C1890 km and the crossrange maneuver zone of 210?C220 km on one side. The maximum guidance error at the altitude of 1 km does not exceed 0.8 km, and the average error is 0.1 km. The algorithm uses up to three bank angle reversals and uses the numerical prediction of the remaining part of the trajectory when choosing the guidance parameters. The concept of the automatic adjustment of the bank angle reference function depending on the entry angle and the prescribed distance to the target point in the maneuver zone is formulated.     

Hierarchical streamline bundles

Effective 3D streamline placement and visualization play an essential role in many science and engineering disciplines. The main challenge for effective streamline visualization lies in seed placement, i.e., where to drop seeds and how many seeds should be placed. Seeding too many or too few streamlines may not reveal flow features and patterns either because it easily leads to visual clutter in rendering or it conveys little information about the flow field. Not only does the number of streamlines placed matter, their spatial relationships also play a key role in understanding the flow field. Therefore, effective flow visualization requires the streamlines to be placed in the right place and in the right amount. This paper introduces hierarchical streamline bundles, a novel approach to simplifying and visualizing 3D flow fields defined on regular grids. By placing seeds and generating streamlines according to flow saliency, we produce a set of streamlines that captures important flow features near critical points without enforcing the dense seeding condition. We group spatially neighboring and geometrically similar streamlines to construct a hierarchy from which we extract streamline bundles at different levels of detail. Streamline bundles highlight multiscale flow features and patterns through clustered yet not cluttered display. This selective visualization strategy effectively reduces visual clutter while accentuating visual foci, and therefore is able to convey the desired insight into the flow data.     

Interactive image segmentation based on level sets of probabilities

In this paper, we present a robust and accurate algorithm for interactive image segmentation. The level set method is clearly advantageous for image objects with a complex topology and fragmented appearance. Our method integrates discriminative classification models and distance transforms with the level set method to avoid local minima and better snap to true object boundaries. The level set function approximates a transformed version of pixelwise posterior probabilities of being part of a target object. The evolution of its zero level set is driven by three force terms, region force, edge field force, and curvature force. These forces are based on a probabilistic classifier and an unsigned distance transform of salient edges. We further propose a technique that improves the performance of both the probabilistic classifier and the level set method over multiple passes. It makes the final object segmentation less sensitive to user interactions. Experiments and comparisons demonstrate the effectiveness of our method.