Collaborative Research: Enabling Discovery in High Reynolds Number Turbulence via Advanced Tools for Petascale SImulation and Analysis

建议没有。项目编号：0749223/0749209/0749235/0749286首席研究员：P-K yeung这项研究将充分利用新兴的PETASCALE计算能力来解决许多重要的研究问题，同时为CFD中的开源代码开发设定新的标准，从而推动高雷诺数湍流科学的发展。科学重点是在最精细的网格分辨率和尽可能高的雷诺数下，模拟均匀湍流和具有一个空间非均匀性方向的非均匀湍流。高级计算的要素将包括领域分解技术，该技术可扩展到未来的千兆级系统，拥有100万个处理器或更多，利用先进硬件特性的高节点级性能，以及增强的存储和分析超大型数据集的能力。将向研究界提供对代码和数据的开放访问。湍流的特点是在时间和空间的大范围尺度上的无序波动，是一个非常复杂和具有社会和技术重要性的问题。根据精确的守恒方程计算波动的直接数值模拟（DNS）是千兆级计算的理想应用，因为需要这种复杂性的计算来解决大范围的空间和时间尺度，而且DNS数据的高可靠性使这种资源投资值得。为了实现PetaScale DNS (PSDNS)，将开发一套功能强大、灵活且可扩展的开源软件来分析结果数据，用于不超过一个空间非均匀性方向的流。PSDNS套件基于pi开发的高度可扩展组件，将进一步开发极端并行性。新的软件将执行许多高雷诺数DNS，以回答湍流研究中的紧迫问题。这些模拟和分析将在湍流研究的各个领域产生重要的发现，包括湍流弥散的间歇性，壁面湍流的高雷诺数重叠层，以及湍流反应流的局部消光和重燃。这项研究将通过湍流研究和计算科学的进步产生广泛的社会和经济影响。空前雷诺数的DNS将影响科学、工程、社会和竞争力等领域，如污染物的混合和扩散、交通工具的设计和阻力、燃烧过程的效率和污染。这项活动还将通过开发基于这些Petascale软件开发的材料，影响高性能计算方面的教育。它将通过从模拟中开发的材料影响流体力学和湍流的教育。最后，在执行所有这些工作的同时，将鼓励代表性不足的群体在各级参与。

PROPOSAL NO.: OCI - 0749223/0749209/0749235/0749286 PRINCIPAL INVESTIGATOR: P-K YEUNGINSTITUTION: Georgia Institute of TechnologyCOLLABORATIVE RESEARCH: ENABLING DISCOVERY IN HIGH REYNOLDS NUMBER TURBULENCE VIA ADVANCED TOOLS FOR PETASCALE SIMULATION AND ANALYSISThis research will advance the science of turbulent fluid flow at high Reynolds number, by taking full advantage of emerging Petascale computing capabilities to address a number of important research questions, while setting a new standard for open-source code development in CFD. The science emphasis is on simulations at the finest grid resolution and highest Reynolds number possible, for homogeneous turbulence and inhomogeneous turbulence with one direction of spatial inhomogeneity. Elements of advanced computing will include domain decomposition techniques that scale to future Petascale systems with on million processors or more, high node-level performance making use of advanced hardware features, and enhanced capacity for storage and analysis of very large datasets. Open access to both codes and data will be provided for the research community. Turbulence is characterized by disorderly fluctuations over a wide range of scales in time and space, and is a problem of great complexity and societal and technological importance. Direct numerical simulations (DNS), in which fluctuations are computed according to exact conservation equations is an ideal application for Petascale computation, since computations of this complexity are needed to resolve the wide range of spatial and temporal scales, and because the high reliability of DNS data makes such a resource investment worthwhile. To enable PetaScale DNS (PSDNS), a powerful, flexible and extensible open-source suite of software analyzing the resulting data, for flows with no more than one direction of spatial inhomogeneity will be developed. The PSDNS suite, based on highly scalable components developed by the PIs, will be further developed for extreme parallelism. New software will perform many high Reynolds number DNS to answer pressing questions in turbulence research. These simulations and analyses will yield critical discoveries in diverse areas of turbulence research, including intermittency in turbulent dispersion, the high Reynolds number overlap layer in wall-turbulence, and local extinction and reignition in turbulent reacting flows. This research will have broad societal and economic impact through advances in turbulence research and computational science. DNS at unprecedented Reynolds numbers will impact science, engineering, society and competitiveness in such areas as mixing and dispersal of pollutants, design and drag of transportation vehicles, and efficiency and pollution in combustion processes. This activity will also impact education in high performance computing through development of materials based on these Petascale software developments. It will impact education in fluid mechanics and turbulence through materials developed from the simulations. Finally, all of this will be performed while encouraging participation at all levels by under-represented groups.