CDS&E: Multi-scale Coherent Structure Extraction and Tracking For Modern CFD Data Analysis

Coherent structures are persistent and recognizable patterns that can be found in fluid flows. In turbulent flows, coherent structures are closely related to a diverse range of physical phenomena, and understanding their behavior is crucial for characterizing, predicting and controlling these flows. However, reliable identification and characterization of coherent structures is challenging due to their diversity and complex inter-relations across different space- and time-scales.

This project brings together experts from both the data visualization and fluid mechanics communities to investigate novel solutions to multi-scale coherent structure extraction, separation, tracking, and visualization. It aims at significantly advancing the ability to analyze large datasets of turbulent flows stemming from computational fluid dynamic (CFD) simulations in a wide range of engineering and scientific applications.

This project provides opportunities for both undergraduate and graduate students with different and diverse backgrounds to participate in the proposed research. The research outcomes can be integrated into the development of a number of undergraduate and graduate courses taught at the University of Houston. The outreach activities enabled by the proposed research help motivate more students to pursue a career in STEM related fields.

To achieve an efficient and reliable analysis for large-scale turbulent flow data, this project aims to investigate a new multi-scale coherent structure representation that encodes relevant flow physics, statistics, and uncertainty information, and to develop a robust computation and exploration framework based on this new representation to support data-driven research. To enable this multi-scale analysis, this project applies a number of spatial and temporal domain decomposition strategies to the computational fluid dynamic (CFD) data.

Multi-field analysis and high-dimensional data projection techniques are adapted to incorporate different physical attributes to the representation. A novel graph representation is leveraged to encode this multifaceted information in a concise and dimension-independent form to enable multi-scale feature extraction and tracking. A matrix representation of this graph is employed to accelerate its processing by utilizing the recent advances in large-scale matrix calculation.

A new visual analytic paradigm is devised based on the proposed graph representation to aid the exploration and comprehension of different turbulence structures individually or collectively. The developed techniques implemented as a number of software libraries can be integrated into existing software, e.g., Paraview, for domain scientists to use in their daily research.

The developed techniques can also be used as pre-processing toolboxes to quantify and extract coherent structures, which can then be visualized by existing software that are not suitable for direct library integration.

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.