Doctor of Philosophy (PhD)
Time-varying data from simulations of dynamical systems are rich in spatio-temporal information. A key challenge is how to analyze such data for extracting useful information from the data and displaying spatially evolving features in the space-time domain of interest. We develop/implement multiple approaches toward visualization-based analysis of time-varying data obtained from two common types of dynamical simulations: molecular dynamics (MD) and computational fluid dynamics (CFD). We also make application case studies. Parallel first-principles molecular dynamics simulations produce massive amounts of time-varying three-dimensional scattered data representing atomic (molecular) configurations for material system being simulated. Rendering the atomic position-time series along with the extracted additional information helps us understand the microscopic processes in complex material system at atomic length and time scales. Radial distribution functions, coordination environments, and clusters are computed and rendered for visualizing structural behavior of the simulated material systems. Atom (particle) trajectories and displacement data are extracted and rendered for visualizing dynamical behavior of the system. While improving our atomistic visualization system to make it versatile, stable and scalable, we focus mainly on atomic trajectories. Trajectory rendering can represent complete simulation information in a single display; however, trajectories get crowded and the associated clutter/occlusion problem becomes serious for even moderate data size. We present and assess various approaches for clutter reduction including constrained rendering, basic and adaptive position merging, and information encoding. Data model with HDF5 and partial I/O, and GLSL shading are adopted to enhance the rendering speed and quality of the trajectories. For applications, a detailed visualization-based analysis is carried out for simulated silicate melts such as model basalt systems. On the other hand, CFD produces temporally and spatially resolved numerical data for fluid systems consisting of a million to tens of millions of cells (mesh points). We implement time surfaces (in particular, evolving surfaces of spheres) for visualizing the vector (flow) field to study the simulated mixing of fluids in the stirred tank.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Bohara, Bidur, "Visualization of Time-Varying Data from Atomistic Simulations and Computational Fluid Dynamics" (2012). LSU Doctoral Dissertations. 977.
Karki, Bijaya B