CDS&E: Data Management and Visualization in Petascale Turbulent Combustion Simulation

CBET-1250171PI: Peyman Givi，匹兹堡大学该奖项为以下方面的发展提供了资金：(1)用于千万亿次湍流燃烧模拟的强大计算方法，(2)用于此类模拟的可扩展数据管理系统，以及(3)生成数据的系统可视化分析。计算方法将基于复杂湍流火焰大涡模拟的滤波密度函数（FDF）方法。模拟将基于一种新的算法，称为？不规则分割拉格朗日蒙特卡罗有限差分？（IPLMCFD），它可以在大规模并行平台上进行FDF模拟，最高可达千兆级平台。数据管理将通过解决注释管理、工作流管理和数据存档方面的研究挑战来提供。有效的数据可视化和分析将通过机器学习、特征提取、的方法。这项工作跨越了工程和计算机科学的学科，扩展了最先进的湍流反应流的高保真预测。在工作结束时，将提供一个开放源代码的LES代码供公众使用。如果成功，这项研究的结果将对燃气轮机工业和政府的燃烧产生重大影响。人们坚信，LES将成为预测未来燃烧系统设计和制造的主要手段。将其与健壮且通用的数据管理和可视化功能相结合，将对基础研究和应用研究都很有用。其他一些更广泛的影响是通过本科生参与研究并吸引他们到研究生院，K-12扩展，高中生参与研究，以及从少数民族和代表性不足的群体招募学生。

CBET-1250171PI: Peyman Givi, Univ. of PittsburghThis award provides funding for developments of (1) a robust computational method for petascale simulationof turbulent combustion, (2) a scalable data management system for such simulations and (3) a systematicvisual analysis of the generated data. The computational method will be based on the filtered density function(FDF) methodology for large eddy simulation (LES) of complex turbulent flames. The simulation willbe based on a new algorithm, termed ?irregularly portioned Lagrangian Monte Carlo-Finite Difference?(IPLMCFD) which facilitate FDF simulations on massively parallel, up to petascale platforms. Data managementwill be provided by addressing research challenges in management of annotations, managementof workflows and data archiving. Effective data visualization and analysis will be conducted through amachine learning ?feature-extraction? approach. This work crosses the disciplines of engineering and computerscience and expands the state-of-the-art in high fidelity predictions of turbulent reacting flows. At theconclusion of the work, an open source LES code will be provided for use by the public.If successful, the results of this research will have a significant impact in combustion, both in gas-turbineindustry and in government. It is firmly believed that LES will constitute the primary means of predictionsfor future design and manufacturing of combustion systems. Having it coupled with robust and versatile datamanagement and visualization capabilities will be useful for both basic and applied research purposes. Someof the other broader impacts are through involvement of undergraduate students in research and attractingthem to graduate school, K-12 outreach, involvement of high school students in research, and recruitmentof students from minority and under-represented groups.