Effect of gas composition on cementite decomposition and coke formation on iron

Cementite decomposition and coke formation in the metal dusting process of iron were investigated at 700 °C in CO-H₂-H₂O gas mixtures. The presence of graphite deposited on the surface initiates the decomposition of cementite into iron and graphite. The morphology of the reaction products varies with gas composition. For CO concentrations less than 5 vol%, particles of iron or even closed iron layers have been observed at the cementite/graphite interface. With increasing CO content the amount of iron in the interface decreases. At CO concentrations higher than 30 vol%, iron could not be detected at the interface by optical microscopy. Thermo-gravimetric analysis shows that the rate of carbon take-up increases with increasing CO concentration reaching a maximum at about 60-75 vol%.The morphologies of graphite in the coke layer can be identified as three types: porous graphite clusters with embedded iron-containing particles, compact bulk graphite with a uniform thickness and a columnar layered structure, and filamentous carbon with iron-containing phases at the tip or along its length. For gas mixtures with low CO concentrations, e.g. 5 vol%, porous graphite clusters are the main form of carbon although filamentous carbon can be seen at the early stage of reaction. With increasing CO concentrations to, e.g. 30 vol%, a compact bulk graphite is formed on the top of the surface. Under this compact graphite, there is an inner layer of graphite which is the combination of porous graphite clusters and filaments. These two layers of graphite are clearly distinguishable when CO content reaches more than 75 vol%. In this case, the main form of graphite in the inner layer is filamentous carbon. The compact graphite layer suffers a serious deformation and forms many cracks because of the growth of catalytic filamentous carbon underneath. These filaments grow outside from compact graphite crevices and finally cover the whole surface. The higher the CO content in the gas, the more the tendency of filamentous carbon formation. The interplay between morphologies of carbon formation and metal dusting has been discussed.     

Effect of copper addition on the thermal contraction of indium melt clusters

The structure of In-1%Cu and In-5%Cu (mass fraction) alloy melts were studied at different temperatures above liquidus by using a high-temperature X-ray diffractometer and were compared with that of pure In melt. Experimental results show that with the addition of 1% Cu or 5% Cu, the thermal contraction phenomenon of atom clusters occurs in melts with the temperature increasing like pure In melt. With the addition of 1% Cu, the thermal contraction of atom clusters increases and the contraction is not homogeneous in the whole measurement temperature range. The sudden change and noticeable contraction can be found in the range of 280-390℃. The temperature range of the sudden change is lower than that of pure In melt. With the addition of 5% Cu, the thermal contraction of atom clusters decreases and the contraction is not consistent in the whole measurement temperature range. The anomalous change can be measured at about 600 ℃. At the same superheating temperature, the nearest interatomic distance r1 of     

基于空间填充曲线的动态负载均衡算法

针对Web集群系统中服务器的数量不断增加、负载指标动态变化的特点,为实现均衡的分配请求,提出一种使用空间填充曲线来实现动态负载均衡的算法。利用空间填充曲线可高效得将高维数据映射到一维索引的特点,使均衡器根据实时收集的各项负载指标快速定位到最优编码的服务器。实验结果表明,该算法能有效地缩短请求响应时间,提升了集群系统的整体性能,在大规模集群系统中均衡效果更好。     

基于聚类和局部区域的彩色图像分割方法

针对现有图像多阈值分割方法存在的分割不够准确、计算复杂度较高等问题,利用Otsu多阈值分割的思想,提出一种基于聚类和局部区域的彩色图像多阈值分割方法.获取彩色图像的H分量直方图;在综合考虑以H分量划分的颜色范围的前提下对直方图中的H分量进行聚类,依次选取各类中的最大值作为对应局部区域的边界;分别搜索局部区域内的最小H分量值,将其作为彩色图像的分割阈值.实验结果表明,与现有方法相比,该方法具有较高的分割准确性和更快的计算速度.     

中医针灸临床治疗专家系统的研究与实现

中医针灸是国之瑰宝,构建中医针灸专家系统是本研究和实践的目标。在《中医针灸信息库》原系统基础上,结合穴位和疗法的有限可枚举性,提出映射样本穴位疗法为二进制串、一种改进的距离函数和一种通过阈值作为收敛标准的K-means聚类算法,为实践目标奠定基础。运用一种评估聚类算法效果设计平衡点的策略,为6万余例针灸治疗案例构建知识库。实践结果表明,该知识库能满足"依病求医"、"依医问病"和知识更新等核心应用需要,查全和查准率好,也能很好地为中医针灸治疗咨询和教学科研提供辅助决策依据。     

基于K-近邻树的离群检测算法

为适应数据集分布形状多样性以及克服数据集密度问题,针对已有算法对离群簇检测效果欠佳的现状,提出了一种基于K-近邻树的离群检测算法KNMOD(outlier detection based on K-nearest neighborhood MST).算法结合密度与方向因素,提出一种基于K-近邻的不相似性度量,然后带约束切割基于此度量构建的最小生成树从而获得离群点.算法可以有效地检测出局部离群点以及局部离群簇,与LOF、COF、KNN及INFLO算法的对比结果也证实了算法的优越性能.     

渲染农场研究进展综述

作为高性能计算领域的新兴应用,渲染有着计算密集和数据密集的应用特征,如何提高渲染农场的效率进行大规模快速渲染,近年来逐渐成为研究的热点问题.对渲染农场的国内外应用现状作了介绍,阐述了渲染农场的组织结构和工作流程.针对渲染农场在优化作业调度算法、改善负载均衡策略以及完善渲染管理平台等三个方面的研究,进行了全面的分析和比较.探讨和总结了在云计算背景下渲染农场在渲染作业调度策略、异构集群渲染任务分发算法、基于云模式的渲染农场新架构和负载均衡策略这四个方面中有待进一步研究的问题.     

面向二进制数据帧的聚类系统

为了分离复杂无线网络环境下获取到的二进制数据帧,为后续协议逆向解析提供前提条件,实现了对复杂协议簇协议的聚类系统.首先使用AC算法挖掘出二进制数据帧中的频繁序列特征;然后创新地使用了Apriori算法搜索分析这些特征的关联关系,并且结合二进制流数据帧的特点对结果进行了四步剪枝处理;最后利用筛选出的特征通过改进的K-means算法进行聚类.实验表明,该系统可以对二进制协议数据帧的聚类起到很好的效果,同时对存在TYPE字段的多层协议簇,还能进一步区分出多种协议间的层次关系.     

W_nH_2(n=1-6)团簇结构与性能的理论研究

采用密度泛函理论中的B3LYP方法在LANL2DZ基组水平上对H_2与W_n(n=1-6)团簇的相互作用进行了计算研究。结果表明,W_nH_2体系的最低能量结构是H2分子的解离性吸附,基本是在W_n团簇最低能量结构的基础上吸附H原子生长而成,说明吸附H原子未改变W_n(n=1-6)团簇的结构和形状;W_nH_2团簇的稳定性比W_n团簇要好;NBO分析表明,W原子比H原子对电荷调节能力稍强,较易与其他原子作用形成新的混合团簇。     

基于MFCC特征提取的故障预测与评价方法

针对渐变故障的准确预测问题,提出了一种基于Mel频率倒谱系数(Mel frequency cestrum coefficient,MFCC)特征提取的故障预测方法.首先通过监测对象上安装的声传感器获得机器当前运行的声音信号;然后声音信号经过预处理后,将其进行Mel变换,获得其MFCC特征向量;最后经SVM分析器对机器运行时的测试数据和所存机器正常运行样本数据进行聚类分析,给出机器的健康评价.测试了海马M3发动机存在撞击声、轻敲声、轰鸣声时的健康度以及分析了采集的声信号存在噪声时的抗噪性能.实验结果表明,MFCC特征值能够作为机器健康预测的特征向量,且方法在预测的准确性和鲁棒性方面具有优势.