SCREMS: Visualization of Complex Spatio-Temporal Multiscale Fluid Dynamic Phenomena

近年来，美国亚利桑那州立大学的流体动力学社区一直在开发强大而高效的流体动力学模拟数值代码，这些数值代码涵盖了不稳定性和湍流过渡的基本问题，以及这些过渡的控制，从几何上简单的几何形状，到微观、中观和全球尺度的大尺度海洋和大气流动模拟，以及开发不同层次相互交织的策略。随着高性能计算计划（HPCI）的建立，亚利桑那州立大学对高性能计算做出了严肃的承诺。虽然我们的计算服务器需求目前由ASU HPCI设施满足，但我们充分利用这一资源的一个严重瓶颈是我们缺乏交互式可视化和分析湍流模拟产生的大型数据集的设施。这可以通过获取交互式可视化集群来解决。将大型复杂时空数据集的交互式图形查询和分析集成到高性能计算范式中，将导致湍流研究方式的转变。可视化集群将使我们能够以一种非常方便的方式，增强传统的流场统计力学分析，这些分析依赖于涉及空间和时间平均值的测量，直接分析湍流连贯结构如何演变和相互作用。对这些4D数据集（三维空间加时间）的分析，不仅包括速度场和温度场，还包括涡度和其他相关的流动量，如螺旋度，目前只能通过大规模的直接数值模拟来实现。对4D模拟数据的可视化查询和分析将在改变我们对复杂时空流的理解方面发挥关键作用。由具有数十年跨学科研究经验的导师对研究生进行跨学科培训是本提案的核心目标。我们的REU参与将由亚利桑那州立大学的csum（数学科学本科生计算科学培训）项目继续进行，在该项目中，本提案中描述的各个单独项目的各个方面将被分离成适合本科生的项目。对于更广泛的社区来说，我们提出的可视化集群以及我们将如何利用它将为大型时空数据集的交互式图形可视化和分析提供一个范例模型。

In recent years, the fluid dynamics community at ASU has been developing robust and efficient numerical codes for fluid dynamics simulations that span from basic fundamental problems concerned with instabilities and transitions to turbulence, and the control of such transitions, in geometrically simple geometries, to large scale oceanographic and atmospheric flow simulations at the micro, meso and global scales and developing strategies to intermesh the different levels. ASU has made a serious commitment to high performance computing with the establishment of the High Performance Computing Initiative (HPCI). While our compute-server needs are presently being met by the ASU HPCI facility, a serious bottleneck to our taking full advantage of this resource is our lack of facilities to interactively visualize and analyze the large data sets that result from our turbulent flow simulations. This is to be addressed with the acquisition of an interactive visualization cluster. The integration of interactive graphical interrogation and analysis of large complex spatio-temporal data sets into the high-performance computing paradigm will lead to a transformation in the way turbulent flows are studied. The visualization cluster will allow us, in a very accessible fashion, to augment the traditional statistical mechanics analyses of the flow fields, which rely on measures involving spatial and temporal averages, with direct analysis of how turbulent coherent structures evolve and interact. The analysis of such 4D data sets (3D spatial plus time) of not only velocity and temperature fields, but also vorticity and other relevant flow quantities such as helicity, is presently only available from large-scale direct numerical simulations. The visual interrogation and analysis of 4D simulation data will play a critical role in transforming our understanding of complex spatio-temporal flows.The interdisciplinary training of graduate students by mentors who have decades of interdisciplinary research experience between them is a core objective of this proposal. Our REU involvement will be continued by the CSUMS (Computational Science Training for Undergraduates in the Mathematical Sciences) program at ASU, where aspects of the various individual projects described in this proposal will be isolated into projects suitable for undergraduates. For the broader community, our proposed visualization cluster and how we will be utilizing it will provide a paradigm model for interactive graphical visualization and analysis of large spatio-temporal data sets.