Collaborative Research: Extreme-scale Ready High-order Methods for Astrophysical and Laboratory Turbulence

合作研究：天体物理和实验室湍流的极端规模就绪高阶方法

• 批准号: 1908834
• 负责人: Dongwook Lee
• 金额: $ 28.37万
• 依托单位: University of California-Santa Cruz
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-08-15 至 2023-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1908834&HistoricalAwards=false
• 关键词: Collaborative; Research; Extreme; scale; Ready

Astrophysical phenomena inherently involve multi-physics and multi-scales that are nonlinearly coupled. The mathematical accuracy and efficiency of the computational tools used to test theoretical models hugely impact the scientific outcomes. The astrophysics community has thus long focused on developing computationally accurate numerical algorithms that give improved solution accuracy, stability, and efficiency. This project tackles the challenge of developing new approaches to address these problems, and incorporating these methods in the FLASH community code in order to reach a large audience in the astrophysics community. The work lies at the nexus of applied and computational mathematics, statistical Gaussian Process (GP) data predictions, and astrophysics. Involvement in NSF-supported outreach will help to educate under-represented community college students in computational astrophysics. Training young scientists in astrophysical and laboratory turbulence using validated simulations also helps to meet a critical national need.This study will develop high-order accurate, novel algorithms with only modest complexity, using a new high-order GP spatial reconstruction approach and a novel single-step high-order temporal integration method. This powerful predictive tool will run on massively parallel high-performance architectures. The study involves, firstly, using the novel GP approach to overcome the second-order bottleneck in most finite-volume shock-capturing astrophysics codes. The GP algorithms provide a simple mathematical framework for computing highly accurate volume-averaged versus pointwise quantities, and multidimensional spatial reconstructions. The second important development is an efficient high-order temporal integration method in adaptive mesh refinement grid configurations. This requires a single-step time update at a single quadrature point per cell face, which will provide the most efficient algorithmic framework in extreme-scale parallel simulations. Finally, comparing simulation results based on solution accuracy, errors, and computational performance for key astrophysical flow problems, including astrophysical and laboratory turbulence, will demonstrate the impact and benefit of these advances.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

天体物理现象本身涉及多物理场和多尺度非线性耦合。 用于测试理论模型的计算工具的数学准确性和效率极大地影响了科学结果。 因此，天体物理学界长期以来一直致力于开发计算精确的数值算法，以提高解的精度、稳定性和效率。 该项目解决了开发新方法来解决这些问题的挑战，并将这些方法纳入FLASH社区代码中，以达到天体物理学社区的大量受众。 这项工作是应用和计算数学，统计高斯过程（GP）数据预测和天体物理学的联系。 参与NSF支持的外展活动将有助于在计算天体物理学方面教育代表性不足的社区大学生。 培训年轻科学家在天体物理和实验室湍流使用验证模拟也有助于满足一个关键的national need.This研究将开发高阶精度，新颖的算法，只有适度的复杂性，使用一个新的高阶GP空间重建方法和一个新的单步高阶时间积分方法。 这个强大的预测工具将在大规模并行高性能架构上运行。 这项研究涉及，首先，使用新的GP方法，以克服二阶瓶颈，在大多数有限体积激波捕捉天体物理代码。 GP算法提供了一个简单的数学框架，用于计算高度准确的体平均量与逐点量以及多维空间重建。 第二个重要的发展是一个有效的高阶时间积分方法在自适应网格细化网格配置。 这需要在每个单元面的单个正交点处进行单步时间更新，这将在极端规模并行模拟中提供最有效的算法框架。 最后，比较模拟结果的基础上解决方案的准确性，误差和计算性能的关键天体物理流问题，包括天体物理和实验室湍流，将证明这些进步的影响和好处。这个奖项反映了NSF的法定使命，并已被认为是值得通过评估使用基金会的智力价值和更广泛的影响审查标准的支持。

项目成果

An Application of Gaussian Process Modeling for High-order Accurate Adaptive Mesh Refinement Prolongation

• DOI: 10.2140/camcos.2022.17.1
• 发表时间: 2020-03
• 期刊: ArXiv
• 作者: Steven I. Reeves;Dongwoo Lee;A. Reyes;C. Graziani;P. Tzeferacos

A Gaussian Process Upsampling Model for Improvements in Optical Character Recognition

用于改进光学字符识别的高斯过程上采样模型

• DOI: 10.1007/978-3-030-64559-5_20
• 发表时间: 2020
• 期刊: International Symposium on Visual Computing
• 作者: Reeves, Steven;Lee, Dongwook;Singh, Anurag;Verma, Kunal
• 通讯作者: Verma, Kunal