基于多GPU的大型线性和非线性方程组的求解(英文)

在处理工程问题时,常常需要对线性或非线性方程组进行求解。对于实际应用中经常遇到的大型方程组进行求解则需要相当长的时间。使用图形处理器(GPU)代替传统的CPU,将多块GPU通过操作系统进行协调,并将PBi-CGstab方法和Inexact Newton方法进行适合多GPU并行的改造以此作为多GPU求解器的核心算法,加速求解大型线性和非线性方程组。本文的多GPU求解器在成倍扩展了单GPU求解器允许的计算规模的同时取得了令人满意的加速比。     

RFID技术在民航领域的应用与展望

射频识别技术(RFID)能够快速、实时、准确、无接触地采集物品信息。随着我国航空业的高速发展,其生产、管理的信息化要求越来越高,RFID技术也开始快速应用于民用航空领域中的飞机部件维修、行李管理等方面。本文简要介绍了RFID在民用航空中的应用状况以及RFID在民用航空中应用会遇到的困难和挑战,提出了相应的解决办法。     

改进遗传算法的纵列式直升机总体参数优化设计

研究纵列式双旋翼直升机总体参数的优化设计方法,在分析双旋翼气动干扰对旋翼气动特性影响的基础上,采用改进的遗传算法,以运输效率为目标、以飞行性能和重量效率等参数为约束条件,建立优化设计模型,优化纵列式直升机总体参数。以CH-47D纵列式直升机为优化算例,计算结果表明,本文提出的优化方法是可行的,优化效果良好。     

民航客机油箱建模与油量传感器姿态误差修正

实现民航客机的飞机油箱建模,提出对油箱的油量传感器示数的姿态误差进行修正的算法.建模过程基于二维限定Delaunay三角剖分和插值,生成油箱的表面.利用油量传感器错误示数和当前飞行姿态角,利用油箱表面三角网格,将油箱内含油体积进行三维三棱柱剖分和三维四面体剖分.计算四面体体积之和求得油箱正确的含油量.与当前正在使用的切片算法在效率和使用范围方面进行深入比较与分析.实现民航客机油箱三维建模,并且利用已建立的模型修正其油量传感器示数姿态误差.算法优于切片算法.     

一种适用于航路改变情况的冲突概率预测算法

针对如何有效地预测飞行中将发生的冲突,提出了适用于航路改变情况下的冲突预测算法.基于飞机实际飞行轨迹的不确定性,计算冲突发生的概率来预测冲突发生的可能性.改进了直线航路预测算法,充分利用ADS-B(广播式自动相关监视)所提供的航路意图信息,获得飞机航路改变点及航路偏转角,依此进行相应的矩阵变换来计算航路改变后的误差协方差矩阵,并利用有效的近似方法对继续累加的误差表示进行简化,进而实现冲突发生概率的计算.运用Monte-Carlo方法进行仿真,结果验证了其较好的准确性,与直线航路算法的比较表明了其在更复杂环境下具有的适用性和有效性.     

Particle Acceleration Mechanisms

In this paper we review the possible mechanisms for production of non-thermal electrons which are responsible for the observed non-thermal radiation in clusters of galaxies. Our primary focus is on non-thermal Bremsstrahlung and inverse Compton scattering, that produce hard X-ray emission. We first give a brief review of acceleration mechanisms and point out that in most astrophysical situations, and in particular for the intracluster medium, shocks, turbulence and plasma waves play a crucial role. We also outline how the effects of the turbulence can be accounted for. Using a generic model for turbulence and acceleration, we then consider two scenarios for production of non-thermal radiation. The first is motivated by the possibility that hard X-ray emission is due to non-thermal Bremsstrahlung by nonrelativistic particles and attempts to produce non-thermal tails by accelerating the electrons from the background plasma with an initial Maxwellian distribution. For acceleration rates smaller than the Coulomb energy loss rate, the effect of energising the plasma is to primarily heat the plasma with little sign of a distinct non-thermal tail. Such tails are discernible only for acceleration rates comparable or larger than the Coulomb loss rate. However, these tails are accompanied by significant heating and they are present for a short time of <10⁶ years, which is also the time that the tail will be thermalised. A longer period of acceleration at such rates will result in a runaway situation with most particles being accelerated to very high energies. These more exact treatments confirm the difficulty with this model, first pointed out by Petrosian (Astrophys. J. 557:560, 2001). Such non-thermal tails, even if possible, can only explain the hard X-ray but not the radio emission which needs GeV or higher energy electrons. For these and for production of hard X-rays by the inverse Compton model, we need the second scenario where there is injection and subsequent acceleration of relativistic electrons. It is shown that a steady state situation, for example arising from secondary electrons produced from cosmic ray proton scattering by background protons, will most likely lead to flatter than required electron spectra or it requires a short escape time of the electrons from the cluster. An episodic injection of relativistic electrons, presumably from galaxies or AGN, and/or episodic generation of turbulence and shocks by mergers can result in an electron spectrum consistent with observations but for only a short period of less than one billion years.     

基于GRNN网络和遗传算法的旋翼动平衡调整

针对传统旋翼调整方法没有考虑调整参数与振动信号之间的非线性关系,提出一种结合广义回归神经网络GRNN(General Regression Neural Network)和遗传算法的旋翼调整方法,采用GRNN建立旋翼动平衡调整模型,以桨叶调整参数作为GRNN输入,以旋翼转轴3个方向的加速度测量值和机身3个方向加速度测量值作为网络输出,建立调整参数与直升机振动信号之间的模型.以直升机振动作为目标函数,采用改进的遗传算法对桨叶调整参数进行寻优,获得直升机振动最小时的桨叶的调整量.飞行实验表明,通过1到2次飞行调整,可使3个方向机身振动(旋翼的一阶振动)为最小,完成旋翼的动平衡调整.      

基于粗神经网络的民用飞机故障诊断

针对传统神经网络故障诊断过程中网络训练时间长、结构复杂以及仅能进行单值输入的缺陷,设计了一种基于粗神经网络的民用飞机故障诊断系统.将粗糙集理论应用在神经网络的前端对民用飞机故障样本数据进行约简处理,以此去除冗余属性的干扰,克服了无关样本数据对网络学习性能的影响,简化了网络结构;利用粗神经元代替传统神经元,提高了网络性能,扩展了网络的应用范围.通过对空中客车A320飞机的故障诊断试验验证了该方法的有效性.     

Report on the Kiso cometary dust trail survey

Cometary dust trails were first observed by IRAS; they are widely known to be the origins of meteoric showers. A new window has been opened for the study of dust trails, using ground-based observations. We succeeded in obtaining direct images of the 22P/Kopff dust trail with the Kiso 1.05-m Schmidt telescope. Following this initial success, we have continued to perform a dust trail survey at Kiso. As a result of this survey, we have detected dust trails along the orbit of six periodic comets, between February 2002 and March 2004. The optical depth of these dust trails are 10⁻⁹ to 10⁻⁸, which is consistent with IRAS measurements. In this paper, we describe the observations and data reduction procedures, and report the brief result obtained between February 2002 and March 2004.     

基于区间灰色不确定语言的工作类型优化方法

针对飞机维修任务工作类型优化过程中的模糊与灰色问题,采用区间灰色不确定语言变量建立工作类型的评价指标和决策模型,提出了军机维修任务工作类型优化方法.对优化方法进行了实例分析,工作类型排序结果和数据对比结果验证了优化方法的可行性与有效性.在某军机上对优化方法进行了工程应用,完成192项初始维修任务工作类型优化,应用结果表...