Fast finite difference Poisson solvers on heterogeneous architectures

In this paper we propose and evaluate a set of new strategies for the solution of three dimensional separable elliptic problems on CPU–GPU platforms. The numerical solution of the system of linear equations arising when discretizing those operators often represents the most time consuming part of larger simulation codes tackling a variety of physical situations. Incompressible fluid flows, electromagnetic problems, heat transfer and solid mechanic simulations are just a few examples of application areas that require efficient solution strategies for this class of problems. GPU computing has emerged as an attractive alternative to conventional CPUs for many scientific applications. High speedups over CPU implementations have been reported and this trend is expected to continue in the future with improved programming support and tighter CPU–GPU integration. These speedups by no means imply that CPU performance is no longer critical. The conventional CPU-control–GPU-compute pattern used in many applications wastes much of CPU’s computational power. Our proposed parallel implementation of a classical cyclic reduction algorithm to tackle the large linear systems arising from the discretized form of the elliptic problem at hand, schedules computing on both the GPU and the CPUs in a cooperative way. The experimental result demonstrates the effectiveness of this approach.     

Melting and reorganization of poly(ethylene terephthalate) on fast heating (1000 K/s)

For poly(ethylene terephthalate) (PET) and other polymers the origin of the multiple melting peaks observed in differential scanning calorimetry (DSC) curves is still controversially discussed. This is due to the difficulty to investigate the melting of the originally formed crystals exclusively. Recrystallization is a fast process and most experimental techniques applied so far do not allow fast heating in order to prevent recrystallization totally. Developments in thin-film (chip) calorimetry allow scanning rates as high as several thousand Kelvin per second. We utilized a chip calorimeter based on a commercially available vacuum gauge, which is operated under non-adiabatic conditions. The calorimeter was used to study the melting of isothermally crystallized PET. Our results on melting at rates as high as 2700 K/s give clear evidence for the validity of a melting-recrystallization-remelting process for PET at low scanning rates (DSC). At isothermal conditions PET forms crystals, which all melt within a few dozens of K slightly above the isothermal crystallization temperature. There is no evidence for the formation of different populations of crystals with significantly different stability (melting temperatures) under isothermal conditions. Superheating of the crystals is of the order of 10 K at 2700 K/s.     

一种快速的判别公共向量分类方法

在传统DCV的基础上,提出了一种改进的快速DCV分类方法。该方法与传统的DCV分类方法相比,在保证识别率相同的情况下具有较快的分类速率。传统的DCV分类方法通过计算特征向量之间的距离来进行分类,而所提快速DCV分类方法则通过标量计算完成分类。理论分析及复杂度计算表明,快速DCV分类方法的分类速率是传统DCV分类方法的2倍左右,在Yale、ORL和PIE 3种人脸数据库得到的对比仿真实验结果验证了该算法的有效性。     

转炉用快速烧结镁r质投补料

根据转炉大面耐火材料的使用条件 ,研究了影响沥青结合及树脂结合镁质投补料使用性能的主要因素。结果表明 ,控制配入合适溶剂的沥青或树脂的加入量 ,掺加少量外加剂 ,可获得性能良好的修补料。二者相比 ,沥青结合镁质投补料的烧结时间较长 ,容易出现凹坑 ;树脂结合投补料与砖体间的粘结性比沥青的稍差 ,容易出现翘曲。     

基于正交试验设计的岩坡力学参数反演

对某岩坡Ⅲ区现场勘察后认为,该区为断层破碎带,属碎裂结构,岩体分级为Ⅳ级.由于地质情况复杂、岩体破碎以及边坡开挖引起的应力释放等原因,致使边坡岩体加固的力学参数用常规方法较难确定.运用正交试验设计的方法借助有限差分程序FLAC进行分析,以安全系数作为评价指标,模拟各参数不同水平的各种组合,对影响边坡稳定性的各因素进行了敏感性分析,并对边坡力学参数进行了反演,得出较为合理的力学参数,为边坡治理提供依据.     

A transient method for characterizing flow regimes in a circulating fluid bed

In recent years, although an increasing number of literature have been devoted to circulating fluidized bed (CFB), the prediction of velocities over which different fluidization regimes exist is still difficult. In this study, a transient method was applied which readily allows one to identify operational features and critical transitions. The method is based on stopping the solids flow rate into the riser when riser is operating in fully dense transport regime. The analysis of transient pressure drop data across the riser during a solids flow cut-off experiment against its time derivative demonstrate the three distinct operating regimes that exist as the gas deplete the solid out of the riser. The transient was compared to data taken under steady state operations using statistically designed experiments. Results indicated that although there were significant differences when comparing operations in dilute conditions, there were no significant differences between the two methods in the fast fluidized and dense transport regimes. The transient method was capable of reproducing the solids circulation dependence on riser solids holdup and on the axial pressure profile. This transient method offers an accurate, easy, rapid, and reproducible means of characterizing CFB operations over a wide range of flow conditions. The lack of accuracy in the dilute regime is conjectured to be due to the wide particle size distribution that resulted in segregation during the transient testing.     

快速直接还原法生产金属化球团工艺的探讨

研究开发了冷固结含碳铁矿球团的还原工艺，它与环形转底炉相结合，形成快速直接还原新工艺，生产金属此球团，具有一系列优点，如：设备简单，生产效率高、能耗低等等。     

基于椭圆曲线密码体系的多基联合稀疏表示算法

椭圆曲线已经广泛应用于密码学领域中的公钥密码系统,数字签名验证等。在椭圆曲线密码系统中,标量乘法占据着十分重要的地位。目前已经有一些算法来加快这些十分关键的计算,它们均基于整数的不同表示,以尽量减少一般乘法或点加法的数目。在椭圆曲线现有的各种计算方法中,联合稀疏表式(JSF),w-NAF,双基方法是目前常用并且有效的算法。基于多基整数表示系统,提出一种联合多基稀疏表示方法。多基联合稀疏表示需要较少的位数,并且比JSF和w-NAF更加稀疏,(2,3,5)基表示与(2,3)基表示相比,可以使椭圆曲线标量乘法的运算提高约16%的效益。     

迭代重加权最小二乘支持向量机快速算法研究

迭代重加权(Iteratively Reweighted)方法是提高最小二乘支持向量机(LS-SVM)稳健性的重要手段,但由于涉及到多次加权和重复训练,该方法需要大量运算,无法广泛应用.通过数值推导,获得了求解迭代重加权最小二乘支持向量机(IRLS-SVM)的快速算法,大幅度减少了其运算复杂度.引入了3种经典的加权函数,并在多个仿真数据集和实际数据集上进行实验,证实了IRLS-SVM能获得相当稳健的学习结果,所提出的快速算法也确实能够大幅度减少训练时间.实验结果同时表明,在快速训练算法的框架下,3种不同的权重函数可能要求不同的训练时间.     

人因操纵特性在航道工程研究中的应用

大型综合航海模拟器在航海教育和培训中扮演着越来越重要的作用。而在某些海事工程研究领域,对于随机生成的复杂环境状况而言,利用航海模拟器进行仿真的方式已经逐渐成为一种非常有效的手段。研究了有关使用大型航海仿真系统航道工程提出了仿真应用方法。对PID控制型船舶自动舵以及船舶在航道航行航迹跟踪应用的研究,提出了一种利用人因操纵特性进行快速仿真的方法,并将其运用于航道工程研究中。