Computation of outflow rates from accretion disks around black holes
人工智能及信息社会-网课答案

如果没找到答案,请关注公众号:搜搜题免费搜题!!!1. 单选题电影()中,机器人最终脱离了人类社会,上演了“出埃及记”一幕。
(1.0分)我,机器人2. 单选题 1977年在斯坦福大学研发的专家系统()是用于地质领域探测矿藏的一个专家系统。
(1.0分)没搜到哦~3. 单选题能够提取出图片边缘特征的网络是()。
(1.0分)卷积层4. 单选题在ε-greedy策略当中,ε的值越大,表示采用随机的一个动作的概率越(),采用当前Q函数值最大的动作的概率越()。
(1.0分)大;小5. 单选题考虑到对称性,井字棋最终局面有()种不相同的可能。
(1.0分)没搜到哦~6. 单选题在语音识别中,按照从微观到宏观的顺序排列正确的是()。
(1.0分)帧-状态-音素-单词7. 单选题没搜到哦~8. 单选题在强化学习过程中,()表示随机地采取某个动作,以便于尝试各种结果;()表示采取当前认为最优的动作,以便于进一步优化评估当前认为最优的动作的值。
(1.0分)探索;开发9. 单选题一个运用二分查找算法的程序的时间复杂度是()。
(1.0分)没搜到哦~10. 单选题典型的“鸡尾酒会”问题中,提取出不同人说话的声音是属于()。
(1.0分)非监督学习11. 单选题 2016年3月,人工智能程序()在韩国首尔以4:1的比分战胜的人类围棋冠军李世石。
(1.0分)AlphaGo12. 单选题首个在新闻报道的翻译质量和准确率上可以比肩人工翻译的翻译系统是()。
(1.0分)微软13. 单选题被誉为计算机科学与人工智能之父的是()。
(1.0分)图灵14. 单选题没搜到哦~15. 单选题科大讯飞目前的主要业务领域是()。
(1.0分)语音识别16. 单选题如果某个隐藏层中存在以下四层,那么其中最接近输出层的是()。
(1.0分)归一化指数层17. 单选题每一次比较都使搜索范围减少一半的方法是()。
(1.0分)没搜到哦~18. 单选题人类对于知识的归纳总是通过()来进行的。
Computational Fluid Dynamics

Computational Fluid Dynamics Computational Fluid Dynamics (CFD) is a branch of fluid mechanics thatutilizes numerical analysis and algorithms to solve and analyze problems that involve fluid flows. It has become an essential tool in various industries, including aerospace, automotive, environmental engineering, and many more. CFD allows engineers and scientists to simulate the behavior of fluids in complex systems, providing valuable insights and predictions that are crucial for design and optimization processes. One of the key challenges in CFD is the accurate modeling of turbulent flows. Turbulence is a complex and chaotic phenomenon that occurs in many practical fluid flow situations. It is characterized by irregular fluctuations in velocity and pressure, making it difficult to predict and analyze using traditional fluid dynamics equations. CFD techniques have been developed to address these challenges, such as large eddy simulation (LES) and detached eddy simulation (DES), which aim to capture the large-scale structures of turbulence while modeling the smaller scales. In addition to turbulence modeling, another significant issue in CFD is the validation and verification of simulation results. Real-world experimental data is often limited, especially in extreme or hazardous environments, making it challenging to validate the accuracy of CFD simulations. Engineers and researchers must carefully validate their CFD models against available experimental data and continuously improve their simulation methodologies to ensure reliability and confidence in the results. Furthermore, the computational cost of CFD simulations can be a significant barrier, especially for complex and large-scale problems. High-fidelity simulations with fine spatial and temporal resolutions can require substantial computational resources and time, limiting the practicality of CFD for some applications. Researchers arecontinually developing and optimizing numerical algorithms and computational techniques to improve the efficiency and scalability of CFD simulations, enabling faster and more cost-effective analyses. From an industry perspective, CFD plays a crucial role in the design and optimization of various engineering systems. In the aerospace industry, CFD is used to analyze airflows over aircraft wings, optimize aerodynamic performance, and reduce drag. In the automotive sector, CFD helps engineers design more efficient and aerodynamic vehicles, leading toimproved fuel efficiency and reduced emissions. Moreover, in the renewable energy sector, CFD is utilized to optimize the design of wind turbines and tidal energy systems, maximizing energy extraction and minimizing environmental impact. Despite its challenges, CFD continues to advance and evolve, driven by the increasing demand for accurate and reliable fluid flow simulations. The ongoing development of high-performance computing, coupled with advancements in numerical methods and turbulence modeling, is pushing the boundaries of what is achievable with CFD. As a result, CFD is poised to play an even more significant role in shaping the future of engineering and technology, offering unprecedented insights into fluid dynamics and empowering engineers to tackle complex design and optimization challenges.。
人工智能题库(附答案)

人工智能题库(附答案)一、单选题(共103题,每题1分,共103分)1.某超市研究销售纪录数据后发现,买啤酒的人很大概率也会购买尿布,这种属于数据挖掘的哪类问题?A、自然语言处理B、分类C、聚类D、关联规则发现正确答案:D2.预测分析过程包括:数据的准备、预测模型开发、模型验收和评估、使用PMML实现大数据预测的有效部署。
()是指对数据的采集和整理A、预测模型开发B、评估C、模型验收D、数据的准备正确答案:D3.Google与Facebook分别提出SimCLR与MoCo两个算法,实现在()上学习图像数据表征。
两个算法背后的框架都是对比学习(contrastivelearning)A、标注数据B、无标注数据C、二维数据D、图像数据正确答案:B4.()是一种模拟人类专家解决领域问题的计算机程序系统。
A、进化算法B、专家系统C、遗传算法D、禁忌搜索正确答案:B5.()的输入为对弈的线路或历史记录,而其输出为目标函数的一系列训练样本。
A、实验生成器B、泛化器C、执行器D、评价器正确答案:D6.话题模型中的几个概念不含有?(___)A、话题B、句C、词D、文档正确答案:B7.主成分分析是一种数据降维和去除相关性的方法,它通过()将向量投影到低维空间。
A、非线性变换B、拉布拉斯变换C、z变换D、线性变换正确答案:D8.查看 Atlas300 (3000)加速卡驱动是否安装成功应该使用哪条命令?A、npusim infoB、npu infoC、atlas-Driver infoD、atlas info正确答案:A9.根据机器智能水平由低到高,正确的是()A、计算智能、感知智能、认知智能B、机器智能、感应智能、认知智能C、机器智能、感知智能、认知智能D、计算智能、感应智能、认知智能正确答案:A10.Python中有这样一个示例:types=['娱乐','体育','科技'],在使用列表时,以下哪个选项,会引起索引错误A、types[0]B、types[-1]C、types[-2]D、types[3]正确答案:D11.剪枝分为前剪枝和后剪枝,前剪枝本质就是早停止,后剪枝通常是通过衡量剪枝后()变化来决定是否剪枝。
2019年度人工智能与健康-公需科目考试及答案79分

分)提出()。
( 2.0 1.1997 年,Hochreiter&Schmidhuber反向传播算法A.深度学习B.博弈论C.长短期记忆模型D.×答错 B 我的答案:分)2.0 2.在大数据隐私保护生命周期模型中,大数据使用的风险是()。
(被第三方偷窥或篡改A.如何确保合适的数据及属性在合适的时间地点给合适的用户访问B.匿名处理后经过数据挖掘仍可被分析出隐私C.如何在发布时去掉用户隐私并保证数据可用D.×答错我的答案: A“先进制造伙伴计划”“人类连接组计划”“创新神经技术脑研究计划”()宣布启动了 3.2.0。
(分)中国A.日本B.美国C.德国D.分)。
( 2.0 4.2005 年,美国一份癌症统计报告表明:在所有死亡原因中,癌症占()A.1/4B.1/3C.2/3D.3/4√答对我的答案:A统计,有()的肿瘤患者需要接受放疗。
WTO 5.癌症的治疗分为手术、放疗、化疗。
据(2.0分)A.18%B.22%C.45%D.70%×答错 A 我的答案:分)2.0 6.到()年,几乎所有的算法都使用了深度学习的方法。
( A.2012B.2014C.2016D.2018×答错我的答案: D分) 2.0 。
(”规划纲要》中提到,健康是经济社会发展的()2030《“健康中国7.A.必然要求基础条件B.核心要义C.根本目的D.8.据《中国心血管病报告2017》(概要)显示,中国现有心血管病患()。
( 2.0分)万人A.1300万人B.1100万人C.450亿人D.2.9√答对 D 我的答案:9.50 年前,人工智能之父们说服了每一个人:“()是智能的钥匙。
”(2.0分)算法A.逻辑B.经验C.学习D.×答错 D 我的答案:10.在()年,AlphaGo 战胜世界围棋冠军李世石。
(2.0分)A.2006B.2012C.2016D.2017√答对 C 我的答案:当前人工智能重点聚焦()大领域。
COMPUTATION OF TURBULENT FLOW IN GENERAL DOMAINS

Abstract
1 Introduction
The purpose of this contribution is to report on experience that has been gathered in the development of a computing method for incompressible ow in complicated domains. We will mainly discuss the path that we have been following towards this goal, and mention alternative approaches only in passing; without any pretension, however, that our way is better. See 14] for a survey of the eld. Our approach may be characterized as a coordinateinvariant generalization of the classical staggered grid discretization in Cartesian coordinates ( 7]) and associated solution methods for incompressible ows.
Supported by the Netherlands Foundation for Mathematics (SMC) with nancial aid from the Netherlands Organization for the Advancement of Scienti c Research (NWO).
影响知识溢出的测度指标研究

影响知识溢出的测度指标研究知识的溢出可以促进企业以较低的成本持续快速的进行创新。
本文通过对影响知识溢出的主要因素进行综合分析,给出了耦合度较小的影响知识溢出的五个指标,为进一步研究知识溢出测度模型提供指标依据。
标签:知识溢出影响因素指标体系知识溢出的存在是科学技术的扩散导致了世界性的技术进步的原因,影响知识溢出的因素有很多而且因素之间关系复杂,使得对于知识溢出的定量分析变的非常困难。
现有关于知识溢出的分析研究大都是针对某一因素进行分析,综合定量分析的较少,定量分析的结果对实际的指导意义有限。
本文通过对知识溢出影响因素的进行综合分析,确定测度知识溢出的指标体系,为知识溢出综合定量分析提供依据。
一、知识溢出的基本原理知识就其本性而言是“非排他的”,但就其产权而言又是“部分排他的”。
一个厂商使用了知识,并不能阻碍其他厂商也使用这一知识。
一旦知识被发现,会立即扩散,并引起经济社会、知识和生产力的进步,但拥有知识的厂商并没有从中获取全部收益。
这种经济的外部性称之为“知识溢出效应”。
知识溢出存在着溢出方、溢出接受方和溢出渠道。
知识溢出方是指向外传播知识的个人、部门、企业以及国家,知识溢出接受方是指接受知识并产生效应的个人、部门、企业以及国家。
在这里我们把知识溢出方和知识溢出接受方统称为个体。
个体之间的知识是可以互动转化的,是一个永无休止的循环过程。
外部知识对个体不会直接产生影响,必须通过与内部知识结合形成个体内部知识才能对个体产生影响。
外来知识进入个体,与个体内部平台知识进行结合,这是一个知识创新的过程,把外部知识内部化,内部化的知识能够对个体产生影响,从而形成知识溢出效应。
知识溢出的过程受到知识溢出方和溢出接受方各自的意愿和能力以及政府政策的影响。
对于溢出方来讲,为了保持自己的竞争优势,防止知识溢出,它可以选择使用不太先进的知识,或者加强对知识的保护,也会督促政府在知识产权保护方面加大力度。
而它的能力便是与其他部门之间的技术差距,借用物理学的概念,可以形象地理解为溢出的“势能”。
河南省三门峡市2024高三冲刺(高考数学)人教版模拟(押题卷)完整试卷
河南省三门峡市2024高三冲刺(高考数学)人教版模拟(押题卷)完整试卷一、单选题:本题共8小题,每小题5分,共40分 (共8题)第(1)题甲、乙两人进行了10轮的投篮练习,每轮各投10个,现将两人每轮投中的个数制成如下折线图:下列说法正确的是()A.甲投中个数的平均数比乙投中个数的平均数小B.甲投中个数的中位数比乙投中个数的中位数小C.甲投中个数的标准差比乙投中个数的标准差小D.甲投中个数的极差比乙投中个数的极差大第(2)题若直线与曲线相切,则()A.B.C.D.第(3)题设,则的大小关系为()A.B.C.D.第(4)题已知集合,集合,则=( )A.B.C.D.第(5)题若复数为纯虚数,则()A.-1B.0C.1D.2第(6)题已知,则()A.B.C.D.第(7)题随着新一代人工智能技术的快速发展和突破,以深度学习计算模式为主的AI算力需求呈指数级增长.现有一台计算机每秒能进行次运算,用它处理一段自然语言的翻译,需要进行次运算,那么处理这段自然语言的翻译所需时间约为(参考数据:,)()A.秒B.秒C.秒D.秒第(8)题若集合{是质数},,则()A.B.C.D.二、多选题:本题共3小题,每小题6分,共18分 (共3题)第(1)题最近几个月,新冠肺炎疫情又出现反复,各学校均加强了疫情防控要求,学生在进校时必须走测温通道,每天早中晚都要进行体温检测并将结果上报主管部门.某班级体温检测员对一周内甲乙两名同学的体温进行了统计,其结果如图所示,则下列结论正确的是()A.甲同学体温的极差为0.4℃B.乙同学体温的众数为36.4℃,中位数与平均数相等C.乙同学的体温比甲同学的体温稳定D.甲同学体温的第60百分位数为36.4℃第(2)题在平面直角坐标系中,已知长为的线段的两个端点和分别在轴和轴上滑动,线段的中点的轨迹为曲线,则下列结论正确的是()A.关于直线对称B.关于原点对称C.点在内D.所围成的图形的面积为第(3)题筒车是我国古代发明的一种水利灌溉工具,因其经济又环保,至今还在农业生产中得到使用(图1),明朝科学家徐光启在《农政全书》中用图画描绘了筒车的工作原理(图2).一半径为2米的筒车水轮如图3所示,水轮圆心O距离水面1米,已知水轮每60秒逆时针匀速转动一圈,如果当水轮上点P从水中浮现时(图中点)开始计时,则()A.点P再次进入水中时用时30秒B.当水轮转动50秒时,点P处于最低点C.当水轮转动150秒时,点P距离水面2米D.点P第二次到达距水面米时用时25秒三、填空题:本题共3小题,每小题5分,共15分 (共3题)第(1)题已知一个圆锥的体积为,其侧面积是底面积的2倍,则其底面半径为_________第(2)题若实数满足约束条件,则目标函数的取值范围是 __________ .第(3)题由正数组成的等比数列中,若,则__________.四、解答题:本题共5小题,每小题15分,最后一题17分,共77分 (共5题)第(1)题已知函数的最小值为6.(1)求的最大值;(2)证明:.第(2)题在中,角所对的边分别为,已知.(1)求角;(2)若为锐角三角形,且,求面积的取值范围.第(3)题已知中,,点在线段上,,.(1)求的大小;(2)求的面积.第(4)题在直角坐标系中,曲线的参数方程为(为参数),以原点为极点,以轴正半轴为极轴建立极坐标系,曲线的极坐标方程为.(1)求曲线的普通方程与曲线的直角坐标方程;(2)已知点是曲线上一个动点,曲线与轴、轴分别交于点,,求面积的取值范围.第(5)题年初新冠病毒疫情爆发,全国范围开展了“停课不停学”的线上教学活动.哈六中数学组积极研讨网上教学策略:先采取甲、乙两套方案教学,并对分别采取两套方案教学的班级的次线上测试成绩进行统计如图所示:(1)请填写下表(要求写出计算过程)平均数方差甲乙(2)从下列三个不同的角度对这次方案选择的结果进行分析:①从平均数和方差相结合看(分析哪种方案的成绩更好);②从折线图上两种方案的走势看(分析哪种方案更有潜力).。
CFD Computations of Emissions for LDI-2 Combustors with Simplex and Airblast Injectors
CFD COMPUTATIONS OF EMISSIONS FOR LDI-2 COMBUSTORSWITH SIMPLEX AND AIRBLAST INJECTORSKumud AjmaniCFD Nexus, LLCCleveland, Ohio, USAHukam Mongia Purdue University West Lafayette, Indiana, USA Phil LeeWoodward FST, IncZeeland, MI, USAABSTRACT An effort was undertaken to perform CFD analysis of fluid flow in Lean-Direct Injection (LDI) combustors with axial swirl-venturi elements for the next-generation LDI-2 design. The National Combustion Code (NCC) was used to perform non-reacting and reacting flow computations on several LDI-2 injector configurations in a thirteen-element injector array for different operating conditions. All computations were performed with a consistent approach of mesh-optimization, spray-modeling, ignition and kinetics-modeling with the NCC. Computational predictions of emissions (EINOx, EICO and UHC) were compared with the two sets of experimental data representing respectively low-pressure and high-pressure engine cycle conditions. INTRODUCTION Emissions targets set for NOx in NASA‘s Environmentally Responsible Aviation program has revived interest in LDI injection concepts and the attendant low emissions levels achieved with these injectors in previous technology demonstration efforts. The results of the previous generation (LDI-1) efforts were summarized by [Lee 2007] and [Tacina 2008], and some highlights of current generation (LDI-2) efforts have been reported by [Lee 2013] and [Suder 2013]. Experimental measurements of axial-swirler LDI-2 configurations based on a combination of thirteen simplex-type pressure-atomizing injectors and airblast injectors as designed by Woodward FST (WFST) were performed at WFST (for low P3) and NASA Glenn Research Center (for high P3). The LDI-2 design used the lessons learnt from experimental studies of nine-element LDI-1 arrays with axial-swirlers, as performed and reported by [Tacina 2005]. In more recent work, [Tacina 2014a] and [Hicks 2014] have reported optical measurements of unsteady effects and comparison of cold and reacting flows, respectively, of LDI-1 multipoint configurations with bladed axial-swirlers and simplex injectors. The National Combustion Code (NCC) has been developed at NASA Glenn Research Center (GRC) through extensive validation with available LDI data, making it very attractive as a tool to help guide technology development of the 2nd generation Lean Direct Injection, LDI-2. Previous work of [Ajmani2013a] and [Mongia, 2008] have shown the importance of establishing CFD best practices for correlating emissions data from state of the art combustions systems. This paper provides an overview of the efforts undertaken with the NCC to compute heat release, NOx and CO emissions (using RANS reacting flow CFD) for LDI-2 designs proposed by Woodward FST, Inc. The CFD results are compared with experimental measurements of NASA GRC [Tacina 2014], and Woodward FST, Inc, at various engine power conditions.COMPUTATIONAL APPROACH WITH THE NCCThe National Combustion Code (NCC) was used to perform simulations of a Woodward, FST, Inc., LDI-2 configuration. The NCC is a state-of-the-art computational tool that is capable of solving the time-dependent, Navier-Stokes equations with chemical reactions. The NCC is being developed primarily at the NASA Glenn in order to support combustion simulations for a wide range of applications, and has been extensively validated and tested for low-speed chemically reacting flows.D o w n l o a d e d b y C R A N F IE L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529 50th AIAA/ASME/SAE/ASEE Joint Propulsion ConferenceJuly 28-30, 2014, Cleveland, OHAIAA 2014-3529Propulsion and Energy ForumThe NCC uses second-order accurate central-differences for the convective and diffusion flux discretization, and a Jameson operator (a blend of 2nd and 4th-order dissipation terms) for numerical stability. The second and fourth order dissipation parameters are typically set to 10-4 and 0.05, respectively ([Swanson 1997]). The value of k 2, the constant that scales the second order dissipation gradient switch, is typically set to 0.25. In order to enhance convergence acceleration in pseudo-time, implicit residual smoothing is used to smooth the computed residuals in NCC RANS. Turbulence closure is obtained by using a two-equation, cubic k-ε model with variable Cµ ([Shih 1998]) and dynamic wall functions with pressure gradient effects ([Shih 2000]).Liquid Phase and Spray ModelingThe specification of the fuel injector exit condition plays a major role in the fidelity of the NCC simulations. Spray injection of fuel particles needs to be specified slightly downstream (1.0mm) of the injector-exit plane of the simplex injectors. In earlier work with NCC RANS, it has been reported that spray initial conditions have a significant impact on reacting-flow results ([Iannetti 2008]). A droplet initial temperature of 300K, and fuel/air ratio dependent initial SMD and injection velocity, as supplied by the injector manufacturer Woodward FST, were specified as inputs for the spray solver. These inputs were based on measurements made by Woodward FST for different fuel-flow rates through the simplex and airblast injectors at simulated atmospheric flow conditions.The best-practice spray inputs were used for the NCC for 60o hollow cone spray with the annular angle width of 100 (viz. droplets spread within 55o to 65o region), 10 droplet groups, discretized into 32 spatial streams along the 3600 circumference. The typical spray integration time-steps were 2e-7s (local time-step, dtml ) and 4e-5s (global time-step, dtgl ), which translates to 200 local time-steps for each global time-step for the spray solver. At each spray time-step, the spatial streams were permitted a stochastic variation of the stream location within the 10o cone thickness. CFD trials with larger number of streams (64, 96) and larger droplet groups (12, 16) with a single element LDI configuration, showed no significant impact on the NCC reacting-flow predictions [Ajmani 2013b].The liquid spray (Jet-A fuel) was modeled by tracking spray particles in a Lagrangian framework, where each particle represents a group of actual spray droplets [Raju 2012]. The governing equations for the liquid phase are based on a Lagrangian formulation where the spray particle position and velocity are described by a set of ordinary differential equations. The Lagrangian solution process used for this study employs a best-practice unsteady spray model such that droplet groups are only integrated for a fraction of their lifetime (but restarted at this point for the next iteration), rather than to a completely steady-state solution. An inflow droplet size distribution is prescribed by the correlation equation: dn n =4.21x 106d d 32!"#$%&3.5e−16.98d d 32()*+,-0.4ddd 32Here n is the total number of droplets, d 32 is the Sauter mean diameter (SMD), and dn is the number of droplets in the size range between d and d + dd. A user-specified number of ‘droplet groups’ is used to represent the drop size distribution among a finite number of droplet classes. As the typical SMD for the conditions computed here was less than 10µm, the specified droplets undergo evaporation without any primary or secondary breakup modeling.Chemical Kinetics and Ignition ModelingA finite-rate chemistry model was used to compute the species source-terms for Jet-A/air chemistry. Reacting flow computations were performed with the chemical-kinetics model described in Table 1. The chemistry model incorporates 14 species and 18 chemical reaction steps. Jet-A fuel is modeled as a surrogate mixture of decane (73%), benzene (18%) and hexane (9%). The kinetics mechanism was validated by matching adiabatic flame temperature, flame-speed and ignition-delay with experimental shock-tube data in the equivalence ratio range of 0.5 to 1.0. The CHEMKIN code was used to find an appropriate activation energy for the fuel breakup step which would produce a close fit between computed D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529ignition delay and experimental ignition delay for a particular equivalence ratio, initial pressure and, initial temperature [Ajmani 2010].The kinetics model uses A (pre-exponential factor), n (temperature exponent) and E (activation energy, cal/mol) to compute the Arrhenius rate coefficient, k = A (T/T 0)n e (-E/RT), for a given temperature, T (K). (R = universal gas constant, T 0 (K) is a reference temperature). Note that reaction steps 1-3 are irreversible, and reaction steps 4-18 are formulated as reversible reactions. The kinetics for NOx prediction includes an extended Zeldovich mechanism (four steps for NO) and an additional four steps for N 2O species. The inclusion of N 2O is expected to improve the NOx predictions in the small local regions where fuel-rich burning is occurring in the flow.Reaction A n E1 C11H21 + O2 => 11CH + 10H + O2 1.00E+12 0.00 3.10E+04GLO / C11H21 0.8/GLO / O2 0.8/2 CH + O2 => CO + OH 2.00E+15 0.00 3.00E+033 CH + O => CO + H 3.00E+12 1.00 0.00E+004 H2 + O2 <=> H2O + O 3.98E+11 1.00 4.80E+045 H2 + O <=> H + OH 3.00E+14 0.00 6.00E+036 H + O2 <=> O + OH 4.00E+14 0.00 1.80E+047 H2O + O2 <=> 2O + H2O 3.17E+12 2.00 1.12E+058 CO + OH <=> CO2 + H 5.51E+07 1.27 -7.58E+029 CO + H2O <=> CO2 + H2 5.50E+04 1.28 -1.00E+0310 CO + H2 + O2 <=> CO2 + H2O 1.60E+14 1.60 1.80E+0411 N + NO <=> N2 + O 3.00E+12 0.30 0.00E+0012 N + O2 <=> NO + O 6.40E+09 1.00 3.17E+0313 N + OH <=> NO + H 6.30E+11 0.50 0.00E+0014 N + N + M <=> N2 + M 2.80E+17 -0.75 0.00E+0015 H + N2O <=> N2 + OH 3.50E+14 0.00 7.55E+0216 N2 + O2 + O <=> N2O + O2 1.00E+15 0.00 3.02E+0217 N2O + O <=> 2NO 1.50E+15 0.00 3.90E+0418 N2O + M <=> N2 + O + M 1.16E+15 0.00 3.32E+0418 N2O + M <=> N2 + O + M 1.16E+15 0.00 3.32E+04Table 1: Kinetics mechanism for Jet-A fuel surrogateA computationally affordable kinetics mechanism (of fewer than 20 species) for liquid spray simulations of Jet-A fuel which can provide NOx and CO predictions, and lean blow-out, at high P 3, high T 3, and low equivalence ratio (< 0.5) conditions for LDI combustor design, remains an open challenge for the chemical kinetics community. Current work at NASA Glenn is focusing on optimizing the current kinetics mechanism for emissions predictions by comparing EINOx and EICO values of the optimized reduced mechanism with emissions values of ‘detailed mechanisms’ and experimental data. More details of the optimization efforts of this kinetics mechanism for emissions predictions for LDI type configurations is provided in [Ajmani 2014].Ignition modeling was performed by introducing artificial ignition source terms in a region 2-3mm downstream of each venturi-exit plane. The ignition sources were turned off in the NCC when every computational cell in the ignition zone reached a ‘cutoff’ temperature of 1600K, or if 1000 iterations with the ignition sources were reached. No “re-ignition” of the mixture was allowed, once the NCC solver had turned off the ignition sources. This practice ensured consistency of CFD computations between the different geometries and different fuel/air ratios computed in this study.D o w n l o a d e d b y C R A N F IE L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529NO Computations and EINOx ComparisonsReacting flow computations with the finite-rate chemistry model for Jet-A/Air described in Table 1 were performed until the mass-weighted averaged EINOx value across the exit plane (located 150mm downstream of the combustor dump plane) was within a range of ±5% of previous values, over 1000 steady-state iterations. A computational averaging of the values of NOx across the entire exit plane was deemed an acceptable best-practice for comparison with experimental data of [Tacina 2014b ].Chemistry-turbulence interaction is an important physical phenomenon to couple the effect of turbulent fluctuations with chemical reactions. The developers of the NCC have implemented and validated several different modeling approaches for chemistry-turbulence interaction. Some results from these efforts have been reported in [Liu 2013]. A computational solution using a Time Filtered Navier Stokes (TFNS) approach with an Eulerian-PDF turbulence-chemistry interaction and detailed chemical kinetics for a single-element LDI configuration was also undertaken and reported in [Ajmani 2013c].LDI-2 GEOMETRY, MESHING AND CFD SETUPThe National Combustor Code (NCC) was used to perform simulations of a multiple elements of the thirteen-element WFST LDI-2 configuration with a combination of airblast and simplex injection elements . A candidate LDI-2 arrangement with no recessed elements (also referred to as ‘baseline’) was chosen for these computations. The Pilot and all three Main stages have their venturi-exit plane in-line with combustor dump-plane. This arrangement, identified as Baseline Config 10, had the following setup for the elements:Pilot: Simplex (pressure atomizer) tip, 55oCCW Main1: 4 elements, Simplex tips, 45o CCWMain2, Main3: 8 Elements, Airblast tips, OAS/IAS 45o CW/45 o CWOAS/IAS – Outer/Inner Air SwirlersCW – Clockwise orientation, CCW – Counter Clockwise orientationThe LDI-2 geometry for the Baseline Config 10 as supplied by WFST (see figure 1) was imported into the CUBIT mesh-generation software, to create a fully tetrahedral mesh with 17M elements. Each blade passage and venturi was meshed as an individual block, and these blocks were then “imprinted” or merged with connecting volumes at their respective common surfaces. This ensured consistency of meshing across similar geometric elements, and also allowed for ‘drop-in’ replacement of single (or multiple) swirlersand/or injectors without needing to regenerate the complete mesh for all thirteen elements.Figure 1. Geometry and mesh for LDI-2 Baseline configuration (Config 10) D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529NCC RANS Computations were performed for three different conditions (as tested at NASA GRC) designated Case A, Case B and Case C, as shown in figure 2. These three conditions may be nominally viewed as low, medium, and high power cycle conditions. Some highlights of the computational analysis with the NCC were:• Fully tetrahedral mesh consists of approximately 17M+ elements. Best-practices for meshing of LDI-1 injector elements reported in [Ajmani, 2013 (ASM)] were leveraged for LDI-2 meshing.• Boundary Conditions– Fixed mass flow rate and static-temperature at inflow– Fixed static-pressure (P 3-dP) at outflow– Adiabatic, no-slip conditions at all wallsPower T 3 avg P 3 avg FAR, P ilot FAR, M ain1 FAR, Main2 FAR, M ain3 FAR, T otal (F) (psi) Case A Low 425 83 0.0654 0.0403 0.0000 0.0000 0.0172 Case B Medium 1001 264 0.0260 0.0261 0.0261 0.0263 0.0251 Case C High 1085 234 0.0257 0.0351 0.0349 0.0352 0.0329dP Air f low AC d P 4 EINOx EIHC EICO T 4 (Expt)(psi) (lb/sec) (in 2) (psi) (g /kg F uel) (g /kg F uel) (g /kg F uel) (F)Case A 2.8 1.0604 1.91 79 2.8 11.2 94.3 1637Case B 10.4 3.0352 2.06 253 3.8 0.0 0.0 2512Case C 7.3 2.3122 2.04 226 11.9 0.0 1.0 2989Figure 2. Three power conditions for Woodward FST LDI-2 design, tested at NASA GRCThe first stage of each simulation consists of non-reacting RANS computations till a ‘mass-imbalance (outflow-inflow) convergence of 0.1% over 500 consecutive NCC RANS iterations is achieved. Typically, 100,000 RANS iterations at a CFL of 1.95 are run, to obtain a converged, steady-state, non-reacting flowfield. The CFL is then lowered to 0.95, and the Lagrangian spray is then initiated for 100 iterations, followed by introduction of artificial ignition sources, to ignite the mixture. An 18 step, 14 species finite-rate chemical kinetics (as discussed earlier in Table 1) is used for the reacting flow simulations. Thereacting RANS computations are run until the value of EINOx at the computational exit plane “converges”. The EINOx is considered converged when its variation over 1000 successive iterations is within a 5% band at the exit plane. For comparison with experimental (five-hole probe) data, computed EINOx values are based on the mass-weighted average of EINOx across the entire exit plane.The specification of liquid fuel flow for each element of the Pilot, Main1, Main2 and Main3 for all cases is listed below:– Each Simplex (Pilot, Main1) injector is modeled with SMD=8.8µm, Vinj=38.6m/s, 60o hollow cone, 8 droplet groups, 32 streams– Each Airblast injector is modeled as 8 or 16 ‘discrete’ circumferential injection holes with SMD=7.5µm, Vinj=5m/s, 10o solid cone, 8 droplet groups, 8 streams. The modeling with 16 ‘discrete’ holes provides better results than 8 holes, as the circumferential film is better resolved with greater number of holes.NCC predictions for effective-area (AC d ) at CASE A conditionsNon-reacting flow: ΔP = 17446Pa = 3.29% of P 3; Computed AC d = 1.98in 2Reacting flow: ΔP = 18908Pa = 3.63%; Computed AC d = 1.90in 2 (Experiment = 1.91in 2) NCC predictions for AC d at Case B conditionsNon-reacting flow: ΔP = 63013Pa = 3.46%; Computed AC d = 2.08in 2Reacting flow: ΔP = 76428Pa = 4.2%; Computed AC d = 1.87in 2 (Experiment = 2.06in 2) NCC predictions for AC d at CASE C conditionsNon-reacting flow: ΔP = 56997Pa = 3.53%; Computed AC d = 1.88in 2Reacting flow: ΔP = 55917Pa = 3.47%; Computed AC d = 1.90in 2 (Experiment = 2.04in 2) D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529Reacting Flow: Axial Velocity Predictions with NCC for Case A (Low Power)Figures 3 and 4 show axial -velocity contour comparisons for NCC predictions for a typical CASE A computation. Cross-sections in four y-planes for the pilot and three Main stages, and six cross-sections in the (axial) x-planes near the dump plane, are plotted in figures 3 and 4, respectively. Figure 3 shows strong primary recirculation zones behind all the pressure atomizer elements, and no recirculation behind any of the air-blast elements. Figure 4 shows that these recirculation regions behind the four pressure atomizers of Main 1 stage are present at the 20mm plane downstream of the dump plane. The experimental measurement plane (114mm) is also shown in figure 4, with a central core of positive axial flow. A five hole-probe was used in the experiment to collect and report emissions data at the 114mm location [Tacina 2014].Figure 3. Axial velocity (m/s) contours for the four planes in line with the Pilot and Main element arrays for Case A.Figure 4. Axial velocity (m/s) contours for six axial planes for Case A; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529Temperature Predictions with the NCC for Case A (Low Power)Figures 5 and 6 show temperature contour comparisons for NCC predictions for Case A. Note that only the Pilot (see Y=0 plane of fig 5) and Main 1 stage (pressure atomizers at Y=0.014 and 0.034) were fueled for this case. The elongated flame behind the pilot (at FAR=0.065) and the shorter flames behind the Main 1 elements (FAR=0.045) are clearly seen in figure 5. The flames for all the PA elements sit in the diverging section of their respective venturi cups. Figure 6 shows that the individual flames from the Main 1 elements start merging with the Pilot flame at 20mm downstream of the dump plane. The 114mm location shows a very distinct hot region along the centerline, and the temperature gradually tapers towards the walls. This trend matches the five-point probe measurements by the traversing probe in the NASA CE5 test cell. The NCC computed mass-averaged value of temperature at the 114mm location is 1082K, as compared to the CEA computed (equilibrium) T 4 of 1165K. The experiment does not measure temperature but uses the CEA computed T 4 (based on FAR, T 3 and P 3) as the ‘experimental’ reference value for T 4.Figure 5. Temperature (K) contours for the four planes in line with the Pilot and Main element arrays for Case A.Figure 6. Temperature (K) contours at six axial planes for Case A; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529NO mass-fraction Predictions with the NCC for Case A (Low Power)Figures 7 and 8 shows NO mass-fraction contour comparisons for NCC predictions for Case A. The NO production from the pilot (Y=0 plane) dominates the overall NO for the system, as seen in figure 7 (Y=0) and the high NO values along the centerline in figure 8 (+10mm, +20mm, +114mm). The 114mm location shows a very high NO concentration at the centerline, which gradually tapers towards the walls. This trend matches the five-point probe measurements by the traversing probe in the experiment. The NCC predicted mass-averaged values of EINOx, EICO and EIHC at the exit are 1.87, 1.7 and 22.8, as compared to the experimental values of 2.81, 94.3 and 11.2, respectively. The EICO prediction is very poor, while the EIHCand EINOx predictions are reasonable.Figure 7. NO mass-fraction contours for the four planes in line with the Pilot and Main element arrays for Case A.Figure 8. NO mass-fraction contours at six axial planes for Case A; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529Reacting Flow: Axial Velocity Predictions with NCC for Case B (Medium Power)Figures 9 and 10 show axial velocity contour comparisons for NCC predictions for a typical Case B computation. Cross-sections in four y-planes for the pilot and three Main stages, and six cross-sections in the (axial) x-planes near the dump plane, are plotted in figures 9 and 10, respectively. Figure 9 shows attached, primary recirculation zones behind all the PA elements (pilot, Main 1), and weak, corner recirculation zones in the venturis of the AB elements (Main 2, Main 3). Figure 10 shows that the recirculation (dark blue) regions behind the Pilot and the four PA elements of Main 1 remain strong at 10mm and 20mm downstream of the dump plane. The experimental measurement plane at 114mm shown in figure 10 shows a fairly well mixed out distribution of axial velocity at this location.Figure 9. Axial velocity (m/s) contours for the four planes in line with the Pilot and Main element arrays for Case B.Figure 10. Axial velocity (m/s) contours for six axial planes for Case B; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529Temperature Predictions with the NCC for Case B (Medium Power)Figures 11 and 12 show temperature contour comparisons for NCC RANS predictions for the Case B computation. The attached flames behind the pressure-atomizing pilot (FAR=0.026) and Main 1 elements (FAR=0.026) are distinct from the weaker, detached flames behind Main 2 and Main 3 airblast elements, as shown in figure 11. Figure 12 shows that the flames from the four Main 1 elements dominate the heat release at the 10mm and 20mm locations. The 114mm location shows some non-mixedness in thetemperature distribution at this location. The computed mass-averaged value of T 4 at the exit plane of the computational domain is 1698K, which is 3% higher than the equilibrium temperature of 1651K computed by the CEA code.Figure 11. Temperature (K) contours for the four planes in line with the Pilot and Main element arrays for Case B.Figure 12. Temperature (K) contours at six axial planes for Case B; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529NO mass-fraction Predictions with the NCC for Case B (Medium Power)Figures 13 and 14 shows NO mass-fraction contour comparisons for NCC RANS predictions for the CaseB conditions. The NO production from pilot element dominates the overall NO for the system, as seen in figure 13 (Y=0.0m, top-left) and figure 14 (see central areas of -10mm, +10mm, +20mm planes). The +20mm location in figure 14 shows that the NCC predicts two high NO regions adjacent to the central pilot. These regions correspond to the Main1 stage, which is fueled by pressure atomizing injectors. The Main2 and Main3 stages, fueled by airblast injectors, produce insignificant amount of NO, even though they are operating at the same FAR as the pilot and Main1 stage. The NCC prediction at 114mm(experimental data location), shows an unmixed profile for NO mass-fraction. The NCC predicted mass-averaged value of EINOx is 5.4, which is 40% higher than the experimentally reported EINOx value of 3.8.Figure 13. NO mass fraction contours for four planes in line with Pilot and Main element arrays (Case B).Figure 14. NO mass fraction contours at six axial planes for Case B; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529Reacting Flow: Axial Velocity Predictions with NCC for CASE C (High Power)Figures 15 and 16 show axial -velocity contour comparisons for NCC predictions for a simulated CASE C condition. Cross-sections in four y-planes for the pilot and three Main stages, and six cross-sections in the (axial) x-planes near the dump plane, are plotted in figures 15 and 16, respectively. Figure 15 shows attached, strong primary recirculation zones behind all the PA elements (pilot, Main 1), and no primary recirculation zones behind the AB elements (Main 2, Main 3). Some ‘corner’ recirculation zones are observed in the diverging sections of the AB venturis. Figure 16 shows that the recirculation (dark blue) regions behind the four PA elements of Main 1 remain strong at +20mm downstream, with a weaker recirculation zone behind the Pilot. The 114mm plane is also shown in figure 16, with a weak central core of positive axial flow surrounded by higher axial velocity near the four Main-3 element locations.Figure 15. Axial velocity (m/s) contours for the four planes in line with the Pilot and Main element arrays for Case C.Figure 16. Axial velocity (m/s) contours at six axial planes for Case C; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529Temperature Predictions with the NCC for CASE C (High Power)Figures 17 and 18 show temperature contour comparisons for NCC predictions for Case C. The attached flames behind the pilot (FAR=0.026) and the much stronger, attached flames behind the Main 1 elements (FAR=0.035) are distinct from the weaker, detached flames behind Main 2 and Main 3 (FAR=0.035) airblast elements (see figure 17). The four Main 1 elements dominate the heat release at the 10mm and 20mm locations downstream of the dump plane, with much smaller heat release from Main2 and Main3 (figure 18). The 114mm location shows four very distinct cold spots near the corners, and considerable non-mixedness in the temperature distribution. The computed mass-averaged value of temperature at the exit is 1922K, which compares excellently with the CEA computed T 4 of 1916K.Figure 17. Temperature (K) contours for the four planes in line with the Pilot and Main element arrays forCase C.Figure 18. Temperature (K) contours at six axial planes for Case C; dump plane is at 0mm. D o w n l o a d e d b y C R A N F I E L D U N I V E R S I T Y o n M a y 11, 2016 | h t t p ://a r c .a i a a .o r g | D O I : 10.2514/6.2014-3529。
数据挖掘智慧树知到课后章节答案2023年下国防科技大学
数据挖掘智慧树知到课后章节答案2023年下国防科技大学国防科技大学绪论单元测试1.什么是KDD? ( )A:领域知识发现B:文档知识发现C:数据挖掘与知识发现D:动态知识发现答案:数据挖掘与知识发现2.“8,000”和“10,000”表示: ( )A:智慧B:知识C:信息D:数据答案:数据3.人从出生到长大的过程中,是如何认识事物的? ( )A:先分类,后聚类B:分类过程C:先聚类,后分类D:聚类过程答案:先聚类,后分类4.“8,000米是飞机飞行最大高度”与“10,000米的高山”表示: ( )A:知识B:数据C:信息D:智慧答案:信息5.“飞机无法飞过高山”表示: ( )A:数据B:信息C:智慧D:知识答案:知识第一章测试1.下面哪个不属于数据的属性类型:( )A:序数B:相异C:区间D:标称答案:相异2.只有非零值才重要的二元属性被称作:( )A:对称属性B:非对称的二元属性C:计数属性D:离散属性答案:非对称的二元属性3.一所大学内的各年纪人数分别为:一年级200人,二年级160人,三年级130人,四年级110人。
则年级属性的众数是: ( )A:三年级B:一年级C:四年级D:二年级答案:一年级4.杰卡德系数用来度量非对称的二进制属性的相似性。
( )A:错 B:对答案:对5.欧式距离用来度量连续数值属性数据的相似性。
( )A:对 B:错答案:对第二章测试1.卡方测试用来度量离散标称属性数据的相关性。
( )A:错 B:对答案:对2.相关系数用来度量标称属性数据的相关性。
( )A:对 B:错答案:错3.所谓高维数据,指的是数据属性很多。
( )A:对 B:错答案:对4.假设属性income的最大最小值分别是12000元和98000元。
利用最大最小规范化的方法将属性的值映射到0至1的范围内。
对属性income的73600元将被转化为:( )A:0.821B:1.458C:0.716D:1.224答案:0.7165.假设12个销售价格记录组已经排序如下:5, 10, 11, 13, 15,35, 50, 55, 72, 92,204, 215 使用如下每种方法将它们划分成四个箱。
2nd IEEE International Conference on Cloud Computing Technology and Science
2nd IEEE International Conference on Cloud Computing Technology and ScienceSelf-Organizing Agents for Service Composition in Cloud Computing (59)J. Octavio Gutierrez-Garcia and Kwang-Mong SimelasticLM: A Novel Approach for Software Licensing in DistributedComputing Infrastructures (67)Claudio Cacciari, Francesco D’Andria, Miriam Gozalo, Björn Hagemeier,Daniel Mallmann, Josep Martrat, David García Peréz, Angela Rumpl,Wolfgang Ziegler, and Csilla ZsigriA Mechanism of Flexible Memory Exchange in Cloud Computing Environments (75)Takeshi Okuda, Eiji Kawai, and Suguru YamaguchiA Novel Approach for Cooperative Overlay-Maintenance in Multi-overlay Environments (81)Chin-Jung Hsu, Wu-Chun Chung, Kuan-Chou Lai, Kuan-Ching Li, and Yeh-Ching ChungFine-Grained Data Access Control Systems with User Accountability in Cloud Computing (89)Jin Li, Gansen Zhao, Xiaofeng Chen, Dongqing Xie, Chunming Rong, Wenjun Li,Lianzhang Tang, and Yong TangTrusted Data Sharing over Untrusted Cloud Storage Providers (97)Gansen Zhao, Chunming Rong, Jin Li, Feng Zhang, and Yong TangA Token-Based Access Control System for RDF Data in the Clouds (104)Arindam Khaled, Mohammad Farhan Husain, Latifur Khan, Kevin W. Hamlen,and Bhavani ThuraisinghamImage Distribution Mechanisms in Large Scale Cloud Providers (112)Romain Wartel, Tony Cass, Belmiro Moreira, Ewan Roche, Manuel Guijarro,Sebastien Goasguen, and Ulrich SchwickerathA Hybrid and Secure Mechanism to Execute Parameter Survey Applicationson Local and Public Cloud Resources (118)Hao Sun and Kento AidaCombinatorial Auction-Based Allocation of Virtual Machine Instancesin Clouds (127)Sharrukh Zaman and Daniel GrosuCost-Optimal Outsourcing of Applications into the Clouds (135)Immanuel Trummer, Frank Leymann, Ralph Mietzner, and Walter BinderTowards a Reference Architecture for Semantically Interoperable Clouds (143)Nikolaos Loutas, Vassilios Peristeras, Thanassis Bouras, Eleni Kamateri,Dimitrios Zeginis, and Konstantinos TarabanisCloudView: Describe and Maintain Resource View in Cloud (151)Dehui Zhou, Liang Zhong, Tianyu Wo, and Junbin KangPerformance Analysis of High Performance Computing Applications onthe Amazon Web Services Cloud (159)Keith R. Jackson, Lavanya Ramakrishnan, Krishna Muriki, Shane Canon,Shreyas Cholia, John Shalf, Harvey J. Wasserman, and Nicholas J. WrightCost-Effective HPC: The Community or the Cloud? (169)Adam G. Carlyle, Stephen L. Harrell, and Preston M. SmithCombining Grid and Cloud Resources by Use of Middleware for SPMD Applications (177)Brian Amedro, Françoise Baude, Fabrice Huet, and Elton MathiasAnalyzing Electroencephalograms Using Cloud Computing Techniques (185)Kathleen Ericson, Shrideep Pallickara, and Charles W. AndersonRapid Processing of Synthetic Seismograms Using Windows Azure Cloud (193)Vedaprakash Subramanian, Liqiang Wang, En-Jui Lee, and Po ChenFinding Tropical Cyclones on a Cloud Computing Cluster: Using ParallelVirtualization for Large-Scale Climate Simulation Analysis (201)D. Hasenkamp, A. Sim, M. Wehner, and K. WuPerforming Large Science Experiments on Azure: Pitfalls and Solutions (209)Wei Lu, Jared Jackson, Jaliya Ekanayake, Roger S. Barga, and Nelson AraujoExploring Architecture Options for a Federated, Cloud-Based System Biology Knowledgebase (218)Ian Gorton, Yan Liu, and Jian YinUsage Patterns to Provision for Scientific Experimentation in Clouds (226)Eran Chinthaka Withana and Beth PlaleSemantics Centric Solutions for Application and Data Portability in Cloud Computing (234)Ajith Ranabahu and Amit ShethAffinity-Aware Dynamic Pinning Scheduling for Virtual Machines (242)Zhi Li, Yuebin Bai, Huiyong Zhang, and Yao MaAchieving High Throughput by Transparent Network Interface Virtualizationon Multi-core Systems (250)Huiyong Zhang, Yuebin Bai, Zhi Li, Niandong Du, and Wentao YangXenrelay: An Efficient Data Transmitting Approach for Tracing Guest Domain (258)Hai Jin, Wenzhi Cao, Pingpeng Yuan, and Xia XiePower-Saving in Large-Scale Storage Systems with Data Migration (266)Koji Hasebe, Tatsuya Niwa, Akiyoshi Sugiki, and Kazuhiko KatoEvaluation and Analysis of GreenHDFS: A Self-Adaptive, Energy-ConservingVariant of the Hadoop Distributed File System (274)Rini T. Kaushik, Milind Bhandarkar, and Klara NahrstedtData Replication and Power Consumption in Data Grids (288)Susan V. Vrbsky, Ming Lei, Karl Smith, and Jeff ByrdResource Provisioning for Enriched Services in Cloud Environment (296)Rosy Aoun, Elias A. Doumith, and Maurice GagnaireUsing Global Behavior Modeling to Improve QoS in Cloud Data Storage Services (304)Jesús Montes, Bogdan Nicolae, Gabriel Antoniu, Alberto Sánchez, and María S. PérezBuilding a Distributed Block Storage System for Cloud Infrastructure (312)Xiaoming Gao, Yu Ma, Marlon Pierce, Mike Lowe, and Geoffrey FoxREMEM: REmote MEMory as Checkpointing Storage (319)Hui Jin, Xian-He Sun, Yong Chen, and Tao KeResource Allocation with a Budget Constraint for Computing IndependentTasks in the Cloud (327)Weiming Shi and Bo HongA Multi-agent Approach for Semantic Resource Allocation (335)Jorge Ejarque, Raül Sirvent, and Rosa M. BadiaInvestigating Business-Driven Cloudburst Schedulers for E-ScienceBag-of-Tasks Applications (343)David Candeia, Ricardo Araújo, Raquel Lopes, and Francisco BrasileiroBag-of-Tasks Scheduling under Budget Constraints (351)Ana-Maria Oprescu and Thilo KielmannA Novel Heuristic-Based Task Selection and Allocation Frameworkin Dynamic Collaborative Cloud Service Platform (360)Biao Song, M.M. Hassan, and Eui-nam HuhCloudBATCH: A Batch Job Queuing System on Clouds with Hadoopand HBase (368)Chen Zhang and Hans De SterckA Novel Parallel Traffic Control Mechanism for Cloud Computing (376)Zheng Li, Nenghai Yu, and Zhuo HaoWork in Progress/Short PapersExploring the Performance Fluctuations of HPC Workloads on Clouds (383)Yaakoub El-Khamra, Hyunjoo Kim, Shantenu Jha, and Manish ParasharScheduling Hadoop Jobs to Meet Deadlines (388)Kamal Kc and Kemafor AnyanwuPetri Net Modeling of the Reconfigurable Protocol Stack for Cloud ComputingControl Systems (393)Hui Chen, Chunjie Zhou, Yuanqing Qin, Art Vandenberg, Athanasios V. Vasilakos,and Naixue XiongTree-Based Consistency Approach for Cloud Databases (401)Md. Ashfakul Islam and Susan V. VrbskyApplication-Oriented Remote Verification Trust Model in Cloud Computing (405)Xiaofei Zhang, Hui Liu, Bin Li, Xing Wang, Haiqiang Chen, and Shizhong WuRecommendations for Virtualization Technologies in High Performance Computing (409)Nathan Regola and Jean-Christophe DucomA Comparison and Critique of Eucalyptus, OpenNebula and Nimbus (417)Peter Sempolinski and Douglas ThainExploratory Project: State of the Cloud, from University of Michiganand Beyond (427)Traci L. RuthkoskiSelf-Caring IT Systems: A Proof-of-Concept Implementation in Virtualized Environments (433)Selvi Kadirvel and José A.B. FortesDynamic Resource Provisioning for Data Streaming Applications in a Cloud Environment (441)Smita Vijayakumar, Qian Zhu, and Gagan AgrawalUser Demand Prediction from Application Usage Pattern in Virtual Smartphone (449)Joon Heo, Kenji Terada, Masashi Toyama, Shunsuke Kurumatani, and Eric Y. ChenForecasting for Grid and Cloud Computing On-Demand Resources Basedon Pattern Matching (456)Eddy Caron, Frédéric Desprez, and Adrian MuresanDynamic Request Allocation and Scheduling for Context Aware ApplicationsSubject to a Percentile Response Time SLA in a Distributed Cloud (464)Keerthana Boloor, Rada Chirkova, Yannis Viniotis, and Tiia SaloInitial Findings for Provisioning Variation in Cloud Computing (473)M. Suhail Rehman and Majd F. SakrVDBench: A Benchmarking Toolkit for Thin-Client Based Virtual Desktop Environments (480)Alex Berryman, Prasad Calyam, Matthew Honigford, and Albert M. LaiAttaching Cloud Storage to a Campus Grid Using Parrot, Chirp, and Hadoop (488)Patrick Donnelly, Peter Bui, and Douglas ThainPower of Clouds in Your Pocket: An Efficient Approach for Cloud MobileHybrid Application Development (496)Ashwin Manjunatha, Ajith Ranabahu, Amit Sheth, and Krishnaprasad ThirunarayanCSAL: A Cloud Storage Abstraction Layer to Enable Portable Cloud Applications (504)Zach Hill and Marty HumphreySustainable Network Resource Management System for Virtual Private Clouds (512)Takahiro Miyamoto, Michiaki Hayashi, and Kosuke NishimuraSafeVanish: An Improved Data Self-Destruction for Protecting Data Privacy (521)Lingfang Zeng, Zhan Shi, Shengjie Xu, and Dan FengBetterLife 2.0: Large-Scale Social Intelligence Reasoning on Cloud (529)Dexter H. Hu, Yinfeng Wang, and Cho-Li WangIntercloud Security Considerations (537)David Bernstein and Deepak VijPerformance Considerations of Data Acquisition in Hadoop System (545)Baodong Jia, Tomasz Wiktor Wlodarczyk, and Chunming RongAbstractions for Loosely-Coupled and Ensemble-Based Simulations on Azure (550)André Luckow and Shantenu JhaResearch Issues for Software Testing in the Cloud (557)Leah Muthoni Riungu, Ossi Taipale, and Kari SmolanderMapReduce in the Clouds for Science (565)Thilina Gunarathne, Tak-Lon Wu, Judy Qiu, and Geoffrey FoxEfficient Metadata Generation to Enable Interactive Data Discoveryover Large-Scale Scientific Data Collections (573)Sangmi Lee Pallickara, Shrideep Pallickara, Milija Zupanski, and Stephen SullivanSpecial Session: Cloud Computing, HCI, & Design--Sustainability and Social ImpactsEnergy Use in the Media Cloud: Behaviour Change, or Technofix? (581)Chris Preist and Paul ShabajeeEnabling Sustainable Clouds via Environmentally Opportunistic Computing (587)Michał Witkowski, Paul Brenner, Ryan Jansen, David B. Go, and Eric WardSocial Impact of Privacy in Cloud Computing (593)Rui Máximo Esteves and Chunming RongOn the Sustainability Impacts of Cloud-Enabled Cyber Physical Space (597)Tomasz Wiktor Wlodarczyk and Chunming RongFraming the Issues of Cloud Computing & Sustainability: A Design Perspective (603)Yue Pan, Siddharth Maini, and Eli BlevisAn Interface Design for Future Cloud-Based Visualization Services (609)Yuzuru Tanahashi, Cheng-Kai Chen, Stéphane Marchesin, and Kwan-Liu MaThe Ethics of Cloud Computing: A Conceptual Review (614)Job Timmermans, Bernd Carsten Stahl, Veikko Ikonen, and Engin BozdagUser Experience and Security in the Cloud – An Empirical Study inthe Finnish Cloud Consortium (621)Nilay Oza, Kaarina Karppinen, and Reijo SavolaCloud Computing for Enhanced Mobile Health Applications (629)M.T. Nkosi and F. MekuriaInternational Workshop on Cloud Privacy, Security, Risk, and TrustOpenPMF SCaaS: Authorization as a Service for Cloud & SOA Applications (634)Ulrich LangSecurity Services Lifecycle Management in On-Demand InfrastructureServices Provisioning (644)Yuri Demchenko, Cees de Laat, Diego R. Lopez, and Joan A. García-EspínModeling the Runtime Integrity of Cloud Servers: A Scoped Invariant Perspective (651)Jinpeng Wei, Calton Pu, Carlos V. Rozas, Anand Rajan, and Feng ZhuInadequacies of Current Risk Controls for the Cloud (659)M. Auty, S. Creese, M. Goldsmith, and P. HopkinsA Privacy Impact Assessment Tool for Cloud Computing (667)David Tancock, Siani Pearson, and Andrew CharlesworthA Framework for Evaluating Clustering Algorithm (677)Joshua Ojo NehinbeDo You Get What You Pay For? Using Proof-of-Work Functions to VerifyPerformance Assertions in the Cloud (687)Falk Koeppe and Joerg SchneiderPrivacy, Security and Trust Issues Arising from Cloud Computing (693)Siani Pearson and Azzedine BenameurCloudSEC: A Cloud Architecture for Composing Collaborative Security Services (703)Jia Xu, Jia Yan, Liang He, Purui Su, and Dengguo FengTrust and Cloud Services - An Interview Study (712)Ilkka Uusitalo, Kaarina Karppinen, Arto Juhola, and Reijo SavolaInternational Workshop on Theory and Practice of MapReduce (MAPRED 2010) HAMA: An Efficient Matrix Computation with the MapReduce Framework (721)Sangwon Seo, Edward J. Yoon, Jaehong Kim, Seongwook Jin, Jin-Soo Kim,and Seungryoul MaengThe Two Quadrillionth Bit of Pi is 0! Distributed Computation of Piwith Apache Hadoop (727)Tsz-Wo SzeHybrid Map Task Scheduling for GPU-Based Heterogeneous Clusters (733)Koichi Shirahata, Hitoshi Sato, and Satoshi MatsuokaPepper: An Elastic Web Server Farm for Cloud Based on Hadoop (741)Subramaniam Krishnan and Jean Christophe CounioCharacterization of Hadoop Jobs Using Unsupervised Learning (748)Sonali Aggarwal, Shashank Phadke, and Milind BhandarkarSSS: An Implementation of Key-Value Store Based MapReduce Framework (754)Hirotaka Ogawa, Hidemoto Nakada, Ryousei Takano, and Tomohiro KudohImplementation and Performance Evaluation of a Hybrid Distributed Systemfor Storing and Processing Images from the Web (762)Murali Krishna, Balaji Kannan, Anand Ramani, and Sriram J. SathishCogset vs. Hadoop: Measurements and Analysis (768)Steffen Viken Valvåg, Dag Johansen, and Åge KvalnesHowdah - A Flexible Pipeline Framework for Analyzing Genomic Data (776)Steven Lewis, Sheila Reynolds, Hector Rovera, Mike O’Leary, Sarah Killcoyne,Ilya Shmulevich, and John BoyleScaling Populations of a Genetic Algorithm for Job Shop Scheduling ProblemsUsing MapReduce (780)Di-Wei Huang and Jimmy LinA Study in Hadoop Streaming with Matlab for NMR Data Processing (786)Kalpa Gunaratna, Paul Anderson, Ajith Ranabahu, and Amit ShethA MapReduce-Based Architecture for Rule Matching in Production System (790)Bin Cao, Jianwei Yin, Qi Zhang, and Yanming YeAuthor Index (796)。
- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
a r X i v :a s t r o -p h /0402555v 1 24 F eb 2004Astronomy &Astrophysics manuscript no.Das February 2,2008(DOI:will be inserted by hand later)Computation of outflow rates from accretion disks around blackholesSantabrata Das,Indranil Chattopadhyay,A.Nandi,and Sandip K.Chakrabarti ⋆S.N.Bose National Centre for Basic Sciences,Salt Lake,Kolkata 700098,IndiaReceived 15th May 2001/Accepted 30th August 2001;Astronomy and Astrophysics,379,683,2001Abstract.We self-consistently estimate the outflow rate from the accretion rates of an accretion disk around a black hole in which both the Keplerian and the sub-Keplerian matter flows simultaneously.While Keplerian matter supplies soft-photons,hot sub-Keplerian matter supplies thermal electrons.The temperature of the hot electrons is decided by the degree of inverse Comptonization of the soft photons.If we consider only thermally-driven flows from the centrifugal pressure-supported boundary layer around a black hole,we find that when the thermal electrons are cooled down,either because of the absence of the boundary layer (low compression ratio),or when the surface of the boundary layer is formed very far away,the outflow rate is negligible.For an intermediate size of this boundary layer the outflow rate is maximal.Since the temperature of the thermal electrons also decides the spectral state of a black hole,we predict that the outflow rate should be directly related to the spectral state.Key words.X-rays:stars –stars:winds,outflows –black hole physics1.IntroductionMost of the galactic black hole candidates are known to undergo spectral state transitions (Tanaka &Lewin,1995;Chakrabarti &Titarchuk,1995,hereafter CT95;Ebisawa et al.1996).Two common states are the so-called hard state and the soft state.In the former,soft-X-ray lu-minosity is low and the energy spectral index α∼0.5(E ν∝ν−α)in the 2-10keV range.In the latter state,the soft-X-ray luminosity is very high,and hard-X-ray inten-sity is negligible.There is also a weak power-law hard-tail component with an energy spectral slope α∼1.5.In the two component advective flow (TCAF)model (CT95),2Santabrata Das et al.:Computation of outflow rates from accretion disks around black holes jets and computed the ratio of the outflow to the inflowrate assuming a simple conical accretion disk model.In the present paper,we compute the absolute value ofthe outflow rate as a function of the rates of the two in-flow components,Keplerian and sub-Keplerian.This wedo analytically following the recently developed proce-dure of obtaining shock locations(Das,Chattopadhyayand Chakrabarti,2001).By dynamically mixing these twocomponents using solutions of the viscous transonicflowswe obtain the specific energy and angular momentum ofthe sub-Keplerian region.We use these pair of parametersto locate shocks in theflow,compute the compression ratioand from this,the outflow rate.We note that as Keplerianmatter is increased in the mixture,the shock compressionratio goes down,and the outflow rate decreases.This isalso the case even from a radiative transfer point of view–when the Keplerian rate is high,the CENBOL region is completely cooled and the shock compression ratio R∼1. Hence in the soft state,which is due to increase of the Keplerian rate,outflow should be negligible.In the next Section,we present the governing equa-tions to compute the outflow rates using a purely analyt-ical method.We compute results for both the isothermal and adiabatic outflows.In§3,we present our results for a single component sub-Keplerianflow.We also produce ex-amples of realistic disks with Keplerian and sub-Keplerian components and obtain outflow rates as functions of the inflow parameters.In§4,we discuss our results and draw conclusions.2.Model EquationsWe consider matter accreting on the equatorial plane of a Schwarzschild black hole.Spacetime around the black hole is described by the Paczy´n ski-Wiita pseudo-Newtonian potentialφ=GM BHdr+1dr−λ22(r−1)2=0.(1) Integrating this,we obtain the conserved specific energy of theflow,E v=12r2−1(γP/ρ)is the adiabatic sound speed.The massflux conservation equation in aflow which is in vertical equilibrium is given by,˙Min=4πρϑrh(r)=Θinρsϑs r2s,(3) whereΘin(= n+14πa s r1/2s)is the solid angle subtended by the inflow at the CENBOL boundary.Subscripts s de-notes the quantities at shock(CENBOL boundary)and h(r)= γar1/2(r−1)is the half-thickness of the disk in vertical equilibrium at a radial distance r.A sub-Keplerianflow with a positive energy will pass through the outer sonic point and depending on whether the Rankine-Hugoniot condition is satisfied or not,a standing shock may form(Chakrabarti,1990;Chakrabarti 1996).If a standing shock forms,then the post-shockSantabrata Das et al.:Computation of outflow rates from accretion disks around black holes3flow would become hotter and would emit hard X-ray ra-diation.This CENBOL region behaves similarly to theboundary of a normal star;it would be expected to driveoutfling Eqs.(2)and(3),it is easy to obtain shocklocations(i.e.,outer surface of the CENBOL)analytically.Briefly,the procedure to obtain shocks involves the follow-ing steps:(a)For a given pair of specific energy E v and angular mo-mentumλ,one obtains a quartic equation for the sonicpoint and solves it for the three sonic points located out-side the horizon.Two of them are saddle type or‘X’typesonic points and one is a centre type or‘O’type sonicpoint.(b)From the inner and the outer‘X’type points,Machnumbers are expressed as polynomials of radial distancer.These Mach number expressions satisfy constraints thatthey must have appropriate values at the sonic points.(c)In addition,it is enforced that the Mach number in-variants at the shock location are also satisfied(r s).(d)The resulting equation becomes quartic in r s and theshock locations are obtained from its solution.Details are discussed in Das et al.(2001).We consideronly the region of the inflow parameter space(E v,λ)thatis able to produce standing shocks.In the pre-shock region,matter is cooler and is sub-Keplerian.Assuming E v∼0(freely falling condition)anda∼0(cool gas)in presence of angular momentum,matterwill fall with a velocity,ϑ(r)= 1r2 1/2.(4)Using this,from Eq.(3)the density distribution can beobtained.At the shock r=r s,i.e.,the boundary of theCENBOL,the compression ratio is given by,R=Σ+h−(r s)ρ−(r s)=ϑ−RΣ−(r s)ϑ2−(r s).(7)The isothermal sound speed in the post-shock regionis obtained from:C2s=W+R2ϑ2−=1r2s(r s−1),(8)where,f0=R24Santabrata Das et al.:Computation of outflow rates from accretion disks around black holescentrifugal barrier and the funnel wall.Generally speak-ing,the outflow surface varies as r 3/2.However,the in-flow surface area is still proportional to r 2.Because of this asymmetry,the problem is no longer tractable ana-lytically and is beyond the scope of the present paper.2.1.When the outflow is isothermalThe radial momentum balance equation in the outflow is given byϑdϑρdP2(r −1)2=0,(9)and the continuity equation is given by1dr(ρϑr 2)=0.(10)Eliminatingdρdr=Nr−1ϑ.To obtain the sonic point condition,we put N =0andD =0and get,ϑ(r c )=C s ,and r c =1+8C 2s ±√8C 2s,where the subscript c denotes the quantities at the sonic point in the outflow.Integrating the radial momentum equation,consider-ing the sonic point condition,we have,C 2s lnρ+−12C 2s +C 2s lnρc −12−1(r s −1)(r c −1).The outflow rate is given by˙M out =Θout ρc ϑc r 2c ,(14)where Θout is the solid angle subtended by the outflow.From Eq.(3)&Eq.(14)we get,˙M out Θinr 2s (r s −1)r s (r s −1)exp [−f ].(15)The above relation is very similar to that obtained in Paper I when the effects of rotation in the inflow were ignored.However,there the ratio R ˙m was a function of R alone.In the present analysis,R is computed self-consistently from the specific energy and the specific an-gular momentum of the flow:R =Σ+ϑ+=112r s=2n +12r c,(17)where the left hand side is the energy at the CENBOL (r =r s )and the right hand side is at the sonic point (r =r c )of the outflow where u c =a c has been used.n =1r s=2n −3f 0−1),(18a )and a 2c4γr c.(18b )In an adiabatic flow with an equation of state P =Kργ(where K is a constant and a measure of entropy),oneobtains,assuming,K c =K s ,ρc a 2sn .(19)Santabrata Das et al.:Computation of outflow rates from accretion disks around black holes5 From these relations one obtains the ratio of the outflowto the inflow rate asR˙m=Θo4γ)3R3[8(R−1)6Santabrata Das et al.:Computation of outflow rates from accretion disks around black holes these shocks is written.Here to compute solid angles of theinflow and the outflow,we assume the half opening angleof the outflow to be10o.Therefore,Θout=π3/162.Θin isgiven in the discussion following Eq.(3).In Paper I,thecompression ratio R was assumed to be a parameter andno angular momentum was assumed a priori.Presently,we show the dependence on angular momentum.The gen-eral character,namely,that the outflow rate is negligiblewhen the shock is weak(R∼1)and falls offgradually forstrongest shock(R→7),remains the same as in Paper I,however.There is a peak at about R˙m∼2.8%.Note thatfor a given R,R˙m increases monotonically with specific an-gular momentumλ.This is because density of CENBOLrises withλ.The curves corresponding toλ=1.71and 1.73are specially marked since there is a clear difference in tendency of the variation of R˙m.For instance,below λ∼1.72,very strong shocks are not possible at all and as a result the outflow rate has a lower limit.Forλ>∼1.72such a limit does not exist.The general behaviour of the outflow rate can be un-derstood in the following way:when shocks are strong, they form very far out,and thus,even though the CENBOL area(which is basically the area of the base of the jet)increases,the net outflow rate is low.When the shock forms very close to the black hole,the temper-ature is high,and thus the outflow velocity is larger,but the CENBOL surface area goes down.Thus the product is low.For the intermediate cases the net effect is larger.For comparison with the analytical work presented in Fig.2a,in Fig.2b we present a similar diagram drawn using a numerical computation of the shock locations (Chakrabarti,1989).Excellent agreement between these twofigures implies that the approximations on which the analytical work was based are justified.All the features are reproduced well in Fig.2a,except that for the weak-est shocks outflow rate is not as low as in the numerical calculation of Fig.2b.We now present the nature of R˙m when the outflow is also chosen to be adiabatic in Fig.3.We usedΘo/Θi∼0.1Fig.2b:Same as Fig.2a except that curves are drawn for the exact numericalsolution.Fig.3:Ratio of the Outflow and the Inflow rates as a function of the compression ratio of the inflow when the outflow is adiabatic.The general nature of the function remains the same as that of the isothermal outflow.for reference.We observe that the peak is still located at around R=∼4and the outflow rate drops for very strong (R∼7)and very weak(R∼1)shocks.We thereforeSantabrata Das et al.:Computation of outflow rates from accretion disks around black holes7 believe that our conclusion about the behaviour of R˙m isgeneric.3.2.Two component advectiveflowsChakrabarti&Titarchuk(1995)proposed that the spec-tral properties are better understood if the disk so-lutions of sub-Keplerianflows are included along withthe Keplerianflows.Recently,Smith,Heindl and Swank(2001),Smith et al.(2001),Miller et al.(2001)found con-clusive evidence of these two components in many of theblack hole candidate accretionflows.While the matterwith higher viscosityflows along the equatorial plane asa Keplerian disk(of rate˙M K),sub-Keplerian halo matter(of rate˙M h)with lower viscosityflanks the Keplerian diskabove and below(Fig.3a).Since the inner boundary con-dition on the horizon forces theflow to be sub-Keplerian,irrespective of their origin(Chakrabarti,1990,1996)mat-ter mixes(at say,r=r tr)from both the Keplerian andsub-Keplerianflows before entering a black hole to forma single component sub-Keplerian with an average energyand angular momentum of E andλrespectively.The spe-cific energy and angular momentum of the mixedflow iscomputed from:E=˙MK E K+˙M h E h˙MK+˙M h.(22)Here,E K,E h,λK andλh are the specific energies and specific angular momentum of the Keplerian and the sub-Keplerian components at r=r tr respectively.Figure4a shows a schematic diagram of the cross-section of a two-component accretionflow.The transition radius(r=r tr)where the Keplerian disk becomes sub-Keplerian,and the shock location r=r s,are indicated. Fig.4b shows two solutions(marked I and II)of the equa-tions governing a two-componentflow(Chakrabarti,1996) whereλd/λK(Sub-Keplerian matter from the Keplerian disk)andλh/λK(Sub-Keplerian halo)are plotted asa Fig.4a:Schematic diagram of the cross section of two-component accretionflow.See text fordetails.Fig.4b:Solution of the two-componentflow equations for two different viscosities.They are merged to form a single solution as depicted in Fig.4a.function of the logarithmic radial distance.Viscosities cho-sen for these two components areα=0.04andα=0.01 respectively.For r<r tr=45(lightly shaded region) the two sub-Keplerianflows mix to create a single com-ponent.For simplicity,we assume viscosity to be negli-8Santabrata Das et al.:Computation of outflow rates from accretion disks around blackholesFig.5:Variation of outflow rates (left axis)with compression ratio at shocks (lower axis).The upper axis gives the variation of sub-Keplerian accretion rate and right axis gives the same for Keplerian accretion rate.gible in this region.Thus,the specific angular momen-tum and specific energy computed at r =r tr from Eqs.(21&22)remain constant (λ)for r <r tr .Dark solid curve (marked III)shows the angular momentum distri-bution λ/λK of all possible mixtures of the two compo-nents which allow shock formation.We chose a case where ˙Md +˙M h =2.0˙M Edd and vary the Keplerian component ˙Md where ˙M Edd is the Eddington accretion rate.In Fig.5,the computed outflow rates are shown when the half opening angle of the outflow is 10o .In this case,Θoutn +1648a s r 1/2s.The left axis shows the rate of out-flow ˙m out =˙M out /˙M Edd as a function of the Keplerian disk rate (right panel)(˙m d =˙M d /˙M Edd )and the halo rate (upper panel)(˙m h =˙M h /˙M Edd ).The lower axis gives the compression ratio at the shock.The most im-portant conclusion that can be drawn here is that the outflow rate steadily goes up as the Keplerian disk rate ˙m d decreases and the spectrum goes to a harder state.When the Keplerian rate is higher,the compression ratio is lower and the outflow rate is also lower.This conclusion,drawn completely from dynamical considerations,is also found to be true from the spectral studies (CT95)where it was shown that the post-shock region cools down and the shock disappears (R →1).Our work therefore hints that the outflow would be negligible in softer states.4.Discussion and Concluding RemarksCT95pointed out that the centrifugal pressure-supported boundary layer (CENBOL)of a black hole accretion flow is responsible for the spectral properties of a black hole candidate.In this Paper ,we present analytical results to show that this CENBOL is also responsible for the pro-duction of the outflows,and the outflow rate is strongly dependent on the inflow parameters,such as specific en-ergy and angular momentum.We showed that in general,the outflow rate is negligible when the shock is absent and very small when the shock is very strong.In intermediate strength,the outflow rate is the highest.As the specific angular momentum is increased,the outflow rate is also increased.This conclusion is valid when the flow is either isothermal or adiabatic.We also demonstrated how a realistic two-component flow (TCAF)consisting of Keplerian and sub-Keplerian components produces a significant amount of outflow.Since matter close to a black hole is sub-Keplerian by nature,the two components must mix to form a single sub-Keplerian flow which has positive specific energy and almost constant specific angular momentum.We showed that as the Keplerian rate of the disk is increased,the out-flow rate is decreased as the shock compression ratio ap-proaches unity.This conclusion,drawn from a dynamical point of view,is also corroborated by the spectral behavior as well —as the Keplerian rate is raised,the post-shock region is cooled due to inverse Comptonization and the shock disappears.This reduces the thermal pressure drive and the resulting outflow rate is reduced.Santabrata Das et al.:Computation of outflow rates from accretion disks around black holes9 Acknowledgements.This work is partly supported by a project(Grant No.SP/S2/K-14/98)funded by Department of Scienceand Technology(DST),Govt.of India.ReferencesChakrabarti,S.K.,1986,ApJ,303,582Chakrabarti,S.K.,1989,ApJ,347,365(C89)Chakrabarti,S.K.,1990,Theory of Transonic AstrophysicalFlows(Singapore:World Sci.)(C90)Chakrabarti,S.K.,Titarchuk,L.G.,1995,ApJ455,623Chakrabarti,S.K.,1996,ApJ464,664.Das,T.K.,Chakrabarti,S.K.1999,Class Quant.Grav,16,3879Das,S.,Chattopadhayay,I.,Chakrabarti,S.K.2001,ApJ,v.557.No.2Ebisawa,K.et al.1996,PASJ,48,59Miller,J.M.,et al.2001,ApJ,submitted(astro-ph/0107514)Paczy´n ski,B.,Wiita,P.J.,1980,A&A,88,23Sakurai,T.1985,Astron.Astrophys.152,121Smith,D.M.,Heindl,W.A.,Swank,J.H.,2001,ApJ,submitted(astro-ph/0103304)Smith,D.M.,Heindl,W.A.,Markwardt,C.,Swank,J.H.,2001,ApJ,554,L41Tanaka,Y.,Lewin,W.H.G.1985,in:(eds.)Lewin.,W.H.G.,Jan Van Paradijs,and van den Heuvel,X-ray Binaries,Cambridge University Press,p.166。