CDS&E: Appraisal of Subgrid Scale Closures in Reacting Turbulence via DNS Big Data

湍流燃烧数据的直接数值模拟（DNS）有助于工业和政府应用的先进燃烧系统的设计和制造。这样的“大数据”集是如此之大，以至于数据要么必须在源头进行严格的过滤，要么在短时间内被丢弃。该项目采用了一系列策略和计算工具，利用DNS数据来评估湍流燃烧中大涡模拟（LES）预测的性能。这项研究将为LES成为预测未来燃烧系统设计和制造的主要手段铺平道路，同时建立数据共享基础设施，并为各级学生提供教育和推广计划。提出的研究是围绕一个协调的5元素策略来处理数十到数百tb大小的湍流燃烧直接数值模拟（DNS）数据集。内容包括：(1)利用不同火焰状态下的DNS数据评估当前的LES策略；(2) LES中SGS关闭置信区间的评估；(3)开发高效计算过滤DNS数据的计算框架；(4)开发基础设施，广泛共享DNS数据和注释，可用于评估未来的SGS关闭和LES预测；(5)对未来在其他（缺失）地区进行火焰DNS的建议。DNS大数据将从多个来源收集，并将属于非预混和预混（全部或部分）火焰。LES将在适用于DNS中考虑的每种火焰配置的子网格规模（SGS）闭包的帮助下进行。将尝试涵盖文献中确定的所有湍流燃烧状态，并进一步了解在不同状态下哪种LES预测效果更好。通过DNS数据对SGS封包进行评估，对于评估封包的信任和信心水平是非常宝贵的。通过整合工程师、计算机科学家和数学家团队的专业知识，这项研究有可能对最先进的高保真湍流燃烧预测产生重大影响。这项研究的成功将对燃气轮机工业和政府（DoD， DOE， NASA）的燃烧产生重大影响。增强的基础设施将有助于LES成为未来燃烧系统设计和制造的主要预测工具，这将促进未来SGS封闭系统的整合。该研究还将为研究生和本科生提供研究机会，扩展K-12课程，并招募少数民族和代表性不足的群体的学生。该项目由计算数据支持科学与工程（CDS&amp；E）计划共同资助。

Design and manufacture of advanced combustion systems for both industrial and government applications is aided by direct numerical simulation (DNS) of turbulent combustion data. Such "big data" sets are so large that the data has to be either aggressively filtered at the source or discarded after a short period of time. The project employs a range of strategies and computational tools for utilizing DNS data to appraise the performance of large eddy simulation (LES) predictions in turbulent combustion. The study will pave the way for LES to become the primary means of predictions for future design and manufacturing of combustion systems, while building a data sharing infrastructure, and providing educational and outreach programs to students at all levels. The proposed research is built around a coordinated 5-element strategy for handling turbulent combustion direct numerical simulation (DNS) data sets of the order of tens to hundreds of terabytes in size. The elements include: (1) Appraisal of current LES strategies using DNS data in various flame regimes; (2) Assessment of confidence intervals of SGS closures in LES; (3) Development of a computational framework for efficient computation of filtered DNS data; (4) Development of infrastructure for broad sharing of DNS data and annotations which can be employed to appraise future SGS closures and LES predictions; and (5) Suggestion for future DNS to be conducted of flames in other (missing) regimes. The DNS big data will be collected from multiple sources and will pertain to both non-premixed and premixed (fully or partially) flames. The LES will be conducted with the aid of subgrid scale (SGS) closures that are applicable for each of the flame configurations considered in DNS. An attempt will be made to cover all of the regimes of turbulent combustion as identified in the literature and contribute further insight as to which LES prediction would work better in the different regimes. Appraisal of the SGS closures via DNS data will be invaluable for assessing the level of trust and confidence that can be placed on the closure. By integrating expertise from a team of engineers, computer scientists, and mathematicians, the study has the potential to make a significant impact in state-of-the-art high-fidelity predictions of turbulent combustion. Success of this research will have a significant impact in combustion, both in the gas-turbine industry and in government (DoD, DOE, NASA). The potential for LES to become the primary predictive tool for future design and manufacturing of combustion systems will be aided by the enhanced infrastructure, which will facilitate incorporation of future SGS closures. The study will also provide research opportunities for both graduate and undergraduate students, K-12 outreach, and recruitment of students from minority and under-represented groups.The project is co-funded by the Computational Data-Enabled Science and Engineering (CDS&E) Program.